Infarcts of Undetermined Cause: The

NINCDS Stroke Data Bank

R. L. Sacco, MD," J. H. Ellenberg, PhD,? J. P. Mohr, MD,' T. K. Tatemichi, MD,' D. B. Hier, MD,$
T. R. Price, MDP and P. A. Wolf, MD"

In a prospective study of 1,805 hospitalized patients in the Stroke Data Bank of the National Institute of Neurological
and CommunicativeDisorders and Stroke, the 1,273 with infarction were classified into diagnostic subtypes. Diagnosis
was based on the clinical history, examination, and laboratory tests including computed tomography, noninvasive
vascular imaging, and where safe and relevant, angiography. Five hundred and eight cases (fully 40%) were labeled as
infarcts of undetermined cause (IUC), of which 138 (27%) were evaluated with both computed tomography and
angiography. The clinical syndrome and computed tomographic and angiographic findings in 91 (65.9%) of these 138
IUC cases were clearly not attributable to largeartery thrombosis and could permit reclassification of the infarct as
due to some form of embolism. Failure to define a source of embolus kept them in the category of IUC. Thirty-one
cases (22.5%) could be reclassified as due to stenosis or thrombosis of a large artery, and 16 (11.6%) as lacunar
infarction. To determine if those selected for angiography among the IUC patients differed from those with other final
diagnoses, a stepwise multiple logistic model was used. The most important characteristics were young age, presence of
a superficial infarct, prior transient ischemic attack, low weakness score, and presentation with a nonlacunar syndrome.
The results of the model suggest that angiography use was determined by clinical characteristics uniformly across
centers and not by final diagnosis. Continued use of the category IUC may help clarify risk factors and stroke subtypes,
allow new mechanisms of ischemic stroke to be uncovered, and prevent classification categories of stroke used in
clinical trials from becoming too broad.
Sacco RL, Ellenberg JH, Mohr JP, Tatemichi TK,Hier DB, Price TR,Wolf PA. Infarcts of undetermined
cause: the NINCDS Stroke Data Bank. Ann Neurol 1989;25:382-390

Many cases of cerebral infarction are difficult to classify
into the traditional diagnostic subgroups of large-vessel
atherothrombosis, small-vessel lacunae, and cardio-
genic embolism. It has been common clinical prac-
tice to classify infarcts into these general categories,
even when the supporting diagnostic data may be in-
adequate. In the National Institute of Neurological
and Communicative Disorders and Stroke (NINCDS)
Stroke Data Bank a classification scheme was devel-
oped to establish uniform diagnostic subgroups based
on supporting data from angiography, brain imaging by
computed tomography (CT) scanning, Doppler ultra-
sonography, clinical syndromes, and related medical
conditions. Patients were classified into a diagnostic
subgroup only after meeting specific criteria. Failure to
satisfy these requirements led to a diagnosis of infarc-
tion of undetermined cause (IUC) in a large number of
patients.
The current emphasis on clinical trials for secondary
stroke prevention calls for a clear understanding of the
diagnosis and mechanism of infarction. The overall size
and uncertain etiological factors of the group currently
known as the IUC group call for a clarification of the
group before it is randomized in treatment trials. This
report describes this large subgroup of patients and
compares their characteristics with the major charac-
teristics of the patients in whom a definitive diagnosis
was made. For the patients labeled as having IUC who
had an angiogram performed, a probable mechanism
of infarction is proposed. The effect of forcible re-
classification of these patients into traditional catego-
ries is demonstrated and the impact this would have on
clinical studies and treatment trials in stroke is dis-
cussed.
Methods
The Stroke Data Bank (SDB), a prospective observational
study, collected acute care and follow-up clinical and labora-

From the *Neurological Institute, Columbia-Presbyterian Medical Received Feb 11, 1988, and in revised form Aug 29. Accepted for
Center, New York, NY;the tBiometry and Field Studies Branch of Dubkation Oct 7. 1988.
iddresscorrespondence to Dr sacco, Neurological Institute, ,
the National Institute of Neurologicai and Communicative Disor-
ders and Stroke, Bethesda, MD; the $Department of Neurology,
West 168th St, Box 6, New York, NY 1o032.
Michael Reese Hospital and Medical Center, Chicago, IL; the 8De-
partment of Neurology, University of Maryland, Baltimore, MD;
and the "Department of Neurology, Boston University Medical Cen-
ter, Boston, MA.

382 Copyright 0 1989 by the American Neurological Association

tory data on cases of stroke. This effort was designed to
provide a resource for research questions on the characteris-
tics, clinical course, and outcome of patients hospitalized
with acute stroke. The collaborative study involved the
Biometry and Field Studies Branch of the NINCDS as the
statistical coordinating center and four academic hospital
centers: University Hospital of Boston University Medical
Center, Boston, MA; Michael Reese Hospital and Medical
Center, Chicago, IL; University of Maryland Hospital, Balti-
more, MD; and the Neurological Institute of Columbia-
Presbyterian Medical Center, New York, NY. A full ac-
count of the design of the SDB can be found elsewhere {I).
Each patient with acute stroke was personally examined by
one of the SDB investigators within a week of onset of
stroke (initial clinical examination at a median of 46 hours
from onset) and underwent initial and subsequent CT scan-
ning (initial CT scan at a median of 20 hours from onset).
A large core of information was collected on each patient
concerning the details of medical, neurological, and social
history; general and neurological examinations; laboratory
studies; and final diagnosis, with special protocols for compli-
cations, stroke evolution, stroke recurrence, and death.
Diagnosis Classification
At the time of hospital discharge, a diagnosis was determined
taking into account all the available data. A classification for
diagnosis was designed to characterize each stroke by causal
mechanism {2}. This classification, used in each clinical cen-
ter, took into account the neurological and medical history,
neurological symptoms and signs, head CT scan, and, when
available, findings from angiography, electrocardiography,
echocardiography, Holter monitoring, and carotid Doppler
ultrasonography.
Strokes were classified into the following categories: in-
farction due to large-artery atherosclerosis (ATH), lacune
(LAC), or embolism from a commonly accepted cardiac
source (EMB); infarction with tandem arterial pathology;
parenchymatous hemorrhage; subarachnoid hemorrhage;
stroke from other unusual causes (OTHER); and infarction
of undetermined cause or infarction with a normal angiogram
(IUC).
A diagnosis of ATH was based on angiographic evidence
of severe stenosis (> 90%) or occlusion of the internal
carotid artery origin or siphon, basilar artery, or major cere-
bral artery stem. A high convexity infarction on CT attrib-
uted to hemodynamic insufficiency also served as evidence of
ATH if it was accompanied by an ipsilateral transient isch-
emic attack (TIA) within the previous 30 days. In the ab-
sence of confirmatory laboratory results, an ipsilateral bruit
or prior TIA may have led to a diagnosis of ATH. Occlusion
without prior TIA was consistent with EMB if associated
with a cardiac source, otherwise it was considered IUC.
LAC was diagnosed in the case of a lacunar syndrome with
a small, deep infarct found on CT or a normal CT scan 1
week following the stroke. A typical lacunar syndrome was
required: pure motor, pure sensory, pure sensorimotor, pure
hemiballism, pure hemichorea, ataxic hemiparesis, or dysar-
thria clumsy hand syndrome. If angiography was performed,
the major ipsilateral cerebral arteries had to be normal. EMB
was diagnosed when a cardiac source was recognized: atrial
fibrillation or flutter, bacterial or marantic endocarditis, mi-
tral annulus calcification, myocardial infarction within the
prior 6 weeks, atrial myxoma, mitral valve prolapse, right-to-
left cardiac shunts, and pulmonary vein thromboses. TAP
represented cases where an extracranial lesion was insuf-
ficient in itself to account for stroke on hemodynamic
grounds, but possibly served as an embolic source. Suppor-
tive findings included a hemispheral surface infarct, a rele-
vant stenosis of more than 75%, a single ulcer more than 2
mm in depth or multiple craters in the internal carotid artery,
and more than 50% stenosis of any major cerebral artery
stem or the basilar artery. The last group, OTHER, was
created to encompass cases of infarction due to arteritis, dis-
section, fibromuscular hyperplasia, or sickle cell anemia;
strokes in a setting of migraine, or mycotic aneurysm; and
other diagnosed but rare or unusual forms of stroke.
The infarct subtype further described in this paper, IUC,
was a diagnosis of exclusion designed to ensure that the
other categories contained a minimum of instances of dubi-
ous or contentious classification. To qualify, patients with
IUC had to have had no bruit or TIA ipsilateral to the
hemisphere affected by the stroke, no obvious cardiac source
of embolism, and a normal CT scan or angiogram or one that
showed definite findings, the interpretation of which did not
allow reliable classification of the type of stroke into one of
the more traditional categories.
Statistical Methods
Because of the importance of angiography in specification of
infarct diagnosis subtype, a special effort was undertaken to
analyze the data to account for the frequency of angiography
use in the data base as a whole and in the IUC subtype in
particular. This effort was important because angiography
was not uniformly applied to all patients with acute stroke.
To assess the uniformity with which angiography was per-
formed, aspects of the clinical history and demographic char-
acteristics of the patient and factors in the initial clinical
evaluation of the presenting stroke were examined to predict
the likelihood of angiography use. A chi-square analysis was
done for each univariate 2 x C contingency table (angiog-
raphy: yes or no versus factor with C levels). Variables
significant at the nominal 0.01 level or clinically suspected to
influence the decision to perform angiography were consid-
ered for multivariate analysis.
The multiple regression approach served as a second
screen to reduce the number of criteria for entry into the
logistic regression analysis, using values of the multiple cor-
relation coefficient and significance levels as guidelines for
selection rather than as strictly valid statistical measures. In-
teraction terms were introduced to account for possible
nonadditive associations of factors with the use of angiog-
raphy. The most important variables from the multiple re-
gression were introduced into a stepwise multiple logistic
model using the stepwise SAS logistic procedure 131.
Limiting the model to a small group of variables, an esti-
mated probability of undergoing angiography was derived
for each patient. The cumulative frequency distributions of
predicted probability were then examined for differences
among centers, C T infarct type, demographic characteristics,
patient outcome, and whether o r not the patient underwent
angiography .
Sacco et al: Infarcts of Undetermined Cause 383
Table I. Distribution of 1,805 Cases in the 1,805
NINCDS Stroke Data Bank All Strokes

No. of Percent of Percent of I

1,273
Stroke Subtype Cases Total Strokes Infarcts Infarcts
I
Infarction 1,273 1,254
Atherosclerosis 113 6.3 8.9 Infarcts With CT Scans
Lacuna 337 18.7 26.5
Cardioembolic 246 13.6 19.3
Tandem arterial 69 3.8 5.4
pathological ab- Abnormal Normal
normality
Infarct of undeter-
mined cause
508 28.1 39.9 505
Related
f i 276
Unrelated Infarct
Hemorrhage Infarct or Other Abnormalitv I
Parenchymatous 237 13.1
hemorrhage
Subarachnoid 243 13.5
hemorrhage
Other 52 2.9
Total 1,805 100.0 100.0 I I I I I I
521155 19/58 671150 1431350 98/218 1251323
NINCDS = National Institute of Neurological and Communicative (34%) (33%) (45%) (41%) (45%) (39%)
Disorders and Stroke.
Fig 1. Proportion in each computed tomography (CT)-angiog-
raphy group with infarcts of undetermined cause (IUC) (138
IUC cases underwent both CT and angiography).
Reckass$cation of IUC Cases
IUC patients were also subdivided based on whether angiog-
raphy was performed, and the group with angiograms was
analyzed separately for probable cause, using the interaction Angiography was carried out to pursue the diagnosis
among clinical syndrome and CT scan and angiogram patho- further in 27% (363 of 1,254) of patients with infarc-
logical findings. tion, yet the final diagnosis was not clarified in a sur-
prising number of patients. Angiograms were obtained
Results on 155 of the 505 whose first CT scan was abnormal
Distribution of Diagnosis Stlbtypes and topographically related to the clinical syndrome. In
Data on 1,805 patients with stroke were available for 34% (52), despite combined findings on the angio-
analysis (Table 1). Of the 1,273 (71%) cases of stroke gram and CT scan, no clear conclusion was reached as
attributed to infarction, the final diagnosis for infarct to the mechanism of the infarct. IUC was also the
subtype showed that atherosclerosis accounted for conclusion by default in 33% (19) of the 58 patients
9%, lacunar infarction for 27%, cardioembolism for who underwent angiography among the 276 whose
1996, and tandem arterial pathology for 5%. Forty initially positive CT scan was clinically unrelated to the
percent (508 cases of infarction) were classified as stroke syndrome. Among the 473 whose first CT scan
IUC, as no proven mechanism was identified despite was normal, 150 patients underwent angiography,
the use of a variety of laboratory tests during hospital- which resulted in a diagnosis no more definitive than
ization. The separate group of patients with stroke IUC in 45% (67 of 150) of the patients. The yield of
from other unusual causes (OTHER), such as dissec- these studies was an average of 38% of cases in which
tion, arteritis, or fibromuscular hyperplasia, accounted IUC was the final conclusion despite the undertaking
for only 2.9% of all patients. of both CT scanning and angiography in pursuit of the
diagnosis. A similar frequency of IUC, approximately
Completeness of Worktlp 41%, was found among the 891 patients whose in-
Data analysis documented that considerable efforts farcts were studied with CT but who did not undergo
were put into the diagnosis of the patients in the SDB. angiography.
At least one adequate CT scan was documented in Although angiography was not performed in 73%
98% (1,254 of 1,273) of the patients with infarction (366) of the 504 CT-scanned IUC patients, other non-
(Fig 1). The first CT scan was normal in 38% (473) and invasive procedures including echocardiography and
abnormal in 62% (781). Infarcts clinically related to carotid Doppler evaluation were pursued in 57%, and
the stroke syndrome were found in 65% (505) of two or more CT scans were obtained in 31% of the
these 781 patients, while 35% (276) had unrelated CT scan-only patients (Fig 2). Despite these extra
infarcts or other clinically irrelevant abnormalities. studies, the diagnosis was no better clarified.

384 Annals of Neurology Vol 25 No 4 April 1989

 INFARCTS OF UNDETERMINED CAUSE
Findings of Tests t o Establish a Diagnosis
508 Although the final diagnosis was IUC, the 138 angio-
grams in these patients revealed a wide variety of
findings (Table 2). Distal intracranial branch occlusions
were observed in 34 patients but were not diagnosed
NO CT CTand CTand as embolism because the algorithm in use required that
No Agrarn Agram No Agram disease be found in the proximal arterial tree or that
4 138 366
(1%) (27%) (72%) cardiac disease be diagnosed by clinical or laboratory
findings. CT scanning in this group disclosed 16
superficial infarcts; 7 superficial and deep infarcts; 3
large, deep infarcts; 7 normal CT scans; and 1 instance
of an isolated small, deep infarct. Using less stringent
Echo'or Echo and Only
Carotid"' Carotid
criteria, all 34 of these patients with branch occlusion
124 85 157 by angiography could have been reclassified as having
(34%) (23%) (43%) been due to embolism of undetermined source.
In 68 patients the angiogram was normal or showed
A
One Twoor unrelated lesions. Of these patients, CT scanning
108 More showed that 8 had superficial infarction, 9 had both
(69Yo) 49 superficial and deep infarction, and 3 had large deep
(31%)
~~ ~ ~~~
infarctions, while 9 showed small deep infarcts, and
Fig 2. Extent of workup in cases of infarcts of undetermined 39 had normal CT scans. The 20 patients who had
cause. * = echocardiogram;** = carotid evaluation included superficial, combined, or large and deep infarctions
directional Doppler ultrasound or duplex Doppler or ocular could also have been classified as having had infarction
plethysmography; C T = computed tomography. due to embolism of undetermined source by less strict

Table 2. Relationship Among Angiogram, CT Scan, and Clinical Syndrome in Infarcts of Undetermined Cause
Clinical Syndrome
Angiographic Finding CT Pathological Finding N Nonlacund Lacunarb
Distal intracranial occlusion Normdunrelated 7 6 1
(n = 34) Superficial infarct 16 15 1
Small, deep infarct 1 1 0
Large, deep infarct 3 3 0
Superficial and deep infarcts 7 7 0
NormaYunrelated (n = 68) Normdunrelated 39 32 7
Superficial infarct 8 8 0
Small, deep infarct 9 6 3
Large, deep infarct 3 2 1
Superficial and deep infarcts 9 9 0
Carotid plaque (n = 5) Normdunrelated 2 2 0
Superficial infarct 0 0 0
Small, deep infarct 0 0 0
Large, deep infarct 1 1 0
Superficial and deep infarcts 2 2 0
Proximal intracranial occlusiod Normdunrelated 4 3 1
stenosis (n = 15) Superficial infarct 4 4 0
Small, deep infarct 3 2 1
Large, deep infarct 1 1 0
Superficial and deep infarcts 3 3 0
Extracranial occlusiodstenosis Normdunrelated 7 6 1
(n = 16) Superficial infarct 0 0 0
Small, deep infarct 2 2 0
Large, deep infarct 0 0 0
Superficial and deep infarcts 7 7 0
Total 138
"Nonlacunar refers to large and small hemispheral, basilar, and other syndromes.
bLacunar refers to pure motor, pure sensory, ataxic hemiparesis, sensorimotor, clumsy hand dysarthria, and hemichoreahallism syndromes.
CT = computed tomography.

Sacco et al: Infarcts of Undetermined Cause 385

algorithms, as could the 32 in whom the CT scan was Table 3. Stroke Diagnosis Subtype of Infarcts
normal in the presence of a nonlacunar syndrome. The of Determined Cause and Reclassifications of Infarcts
9 with a small, deep infarct and the 7 with normal CT of Undetermined Cause with Angiography
~~ ~

scan and a lacunar syndrome that did not fully meet Infarcts of
the SDB diagnostic criteria for LAC could possibly Undetermined Infarcts of
have been reclassified as having had lacunes using less Cause with Determined
stringent definitions. One of the latter 7 had atrial Angiography Cause
Stroke Diagnosis
fibrillation on the admission electrocardiogram. Subtype n Percent n Percent
Five had ipsilateral carotid plaque with computed
Embolism" 91 65.9 308 41.1
tomographic evidence of infarction or a normal CT
Lacune 16 11.6 330 44.0
scan; none had a lacunar syndrome. These cases should
Atherothrombosis 31 22.5 112 14.9
be considered examples of embolism of presumed
Total 138 100.0 750 100.0
carotid source, but were classified as IUC by the ob-
servers because the plaque seen on the angiogram did "Embolism includes cases of cardioembolism (240) and tandem arte-
not appear to be of sufficient size to be responsible for rial pathology (68) in infarcts of determined cause and cases of em-
bolism of undetermined source in infarcts of undetermined cause
the infarct. with angiography.
In 3 1 patients, ipsilateral proximal intracranial or ex- n = number of cases with computed tomography scan.
tracranial stenosis or occlusion was documented. In 24,
the CT scan either showed an infarct that was not a
small, deep infarct ( 1 5 ) or was normal or unrelated in Table 4. Logistic Model for Prediction of Use of Angiographya
the presence of a nonlacunar syndrome (9). In 7, the
All Infarcts (n = 1,254)
large-artery disease could have been responsible for a ~

small-vessel infarction, since CT scans of 5 patients Order of

revealed a small, deep infarct and scans of 2 others Factor Coefficient x2 Entry
revealed a normal or clinically unrelated lesion and a Intercept (a) 3.747
lacunar syndrome. Assuming an occlusion was not due Age - 0.067 123 1
to embolism, a stenosis was not a recanalizing embolus, CT: superficial infarct 0.869 18 2
and a small, deep infarction was the result of penetrat- TIA 0.842 21 3
ing arterial atherosclerosis, the stroke mechanism in Weakness score - 0.500 27 4
these 31 patients could be reclassified as atheroscle- Typical lacunar syndromeb - 0.754 17 5
rosis.
Taking into consideration the clinical syndrome and "Logistic model can be described as Pr(angiogrdfactors) = (1 +
exp - (a + px))-'.
findings on CT scan and angiogram, the infarcts in 91 bExcludes sensorimotor lacune.
(65.9%) of these 138 IUC patients were clearly not CT = computed tomography; TIA = transient ischemic attack.
attributable to large-artery thrombosis and could be
classified as resulting from some form of embolism,
despite the lack of a determined source. Thirty-one 4). Age, weakness score, and lacunar syndrome were
(22.5%) infarcts could be reclassified as having re- inversely related to the likelihood of angiography. Us-
sulted from a stenosis or thrombosis of a large artery, ing the logistic model, patients were assigned a proba-
and 16 (11.6%) from lacunes (Table 3). bility of undergoing angiography. The correlation for
the model (r = .42) should be viewed with caution as a
Predictors of Angiography measure of association 141.
The decision to pursue angiography in an individual There was a modest variation among the centers in
patient was not based on an algorithm; rather it was the proportion of patients with infarction in whom an-
partly dependent on the practice patterns of the pa- giography was performed: Boston 22.296, Chicago
tient's physician. The question posed was whether a 24.796, Maryland 29.6%, New York 35.9%. How-
consistent algorithm for the decision to perform an- ever, the institutional rate of angiography followed the
giography was used in all patients or whether a differ- respective proportion of patients at high probability to
ent algorithm was applied to those whose infarct was undergo angiography. For example, New York had the
finally diagnosed as IUC. Using a multiple logistic pre- highest rate of angiography but also had the .highest
diction model, for all 1,254 cases of infarction that proportion (24%) of patients with a greater than 0.5
were examined by CT at the time of admission the predicted probability for undergoing angiography.
most important predictors of whether angiography was Based on the logistic model, the average predicted
performed were, in order of importance, age at stroke, probabilities of undergoing angiography in each of the
presence of a superficial infarct, prior TIA, weakness centers are: Boston 25.4%, Chicago 26.7%, Maryland
score, and presentation of a lacunar syndrome (Table 34.196, and New York 34.6%.

386 Annals of Neurology Vol 25 No 4 April 1989

Table 5 . Characteristics of Infarcts of Undetermined Cause (IUC) Compared with Those of Infarcts
of Determined Cause (IDC) with and Without Angiography
IUC With IUC Without IDC With IDC Without
Angiography Angiography Angiography Angiograph y
No. of cases with CT scan 138 366 225 525
Age at strokeqb(yr)
Mean 58 71 60 69
Range 17-84 25-99 18-84 19-99
Sex: (%) 56 41 59 45
Race: white (%) 31 29 40 36
Total weakness scoreqb(%)
Mild 59 38 53 41
Moderate 25 24 32 33
Severe 16 38 15 26
Hemispheral syndromeb (%) 65 66 47 38
Basilar syndromeb (%) 15 18 12 5
Classic lacunar syndromeb.c(%) 5 7 24 42
CT: superficial infarctb(%) 20 13 23 8
Worsening during hospitalizationb(%) 27 28 33 17
30-day fatality" (%) 4 14 5 6
Risk factors (a)
Previous stroke 21 28 26 26
Prior T I A ~ , ~ 20 15 30 11
Hypertensiona 62 71 66 70
Myocardial infarctionb 14 19 13 22
Valvular diseaseb 1 3 2 8
Atrial fibrillationb 1 4 1 20
Other arrhythmiab 4 9 5 12
Angina 15 19 17 17
Congestive heart failure*b 5 17 5 20
Prior heart conditionqb*e 21 33 20 46
Diabetes 28 30 24 25
T h e computed chi-square statistic for the difference of the distribution of this characteristicbetween IUC cases with and without angiography
was, with the exception of hypertension ( p = 0.053), less than the tabled value of chi-square at p = 0.01.
"The computed chi-square statistic for the difference of the distribution of this characteristicbetween IDC cases with and without angiography
was, with the exception of hemispheral syndrome ( p = 0.024), less than the tabled value of chi-square at p = 0.01.
'Excludes sensorimotor lacune.
approximately 14% of IUC cases and 8% of IDC cases the frequency of prior TIA was unknown.
'Includes 1 or more of the following conditions: myocardial infarction, valvular heart disease, atrial fibrillation, other arrhythmias, and congestive
heart failure.
CT = computed tomography; TIA = transient ischemic attack.

The predicted probability of undergoing angiog-
raphy was computed for each IUC case. The frequency
distribution of the predicted probabilities for the IUC
cases with angiography was compared with that for the
IUC cases without angiography and was found to be
statistically different (x2 [9] = 96; p < 0.001). The
mean predicted probability was 0.47 and 0.26, respec-
tively, indicating an appropriate discrimination be-
tween those with and without angiography. For the
total cohort of IUC cases, the average predicted proba-
bility for undergoing angiography was 0.32 and the
actual proportion of cases studied with angiography
was 29%. For infarcts of determined cause (IDC;
ATH, EMB, LAC, TAP) the proportion of cases
studied with angiography was 30.6%, and the average
predicted probability was 0.29. These results suggest
that angiography use in patients whose final diagnosis
was IUC was determined similarly to its use in those
whose final diagnosis was IDC.
Comparison of IUC and IDC With
and Without Angiography
Table 5 compares the four groups of IUC and IDC
patients with and without angiography. As indicated by
the model, the patients who underwent angiography in
both groups were younger and were less severely im-
paired based on total weakness scores.
For IDC, the patients who underwent angiography
were more likely to be men and more frequently had a
history of prior TIA, superficial infarcts visualized by
CT, and hemispheral or basilar syndromes, and their
condition was more likely to have worsened in the

Sacco et al: Infarcts of Undetermined Cause 387

hospital. Lacunar syndromes and cardiac risk factors
were more prevalent in the group not studied with
angiography. The prevalence of previous stroke, hy-
pertension, and diabetes, and distribution of race were
similar among the groups.
For IUC, the mean age at stroke in the CT and
angiography subgroup was 58. Twenty-seven percent
worsened in the hospital, and 41% had a moderate to
severe weakness score. Hemispheral syndromes pre-
dominated in 66%; basilar syndromes occurred in
15%. CT demonstrated clinically relevant infarcts in
57%; surface infarction was found in 40%.
The IUC patients without angiography differed
from those who underwent angiography in being older
and having a higher frequency of hypertension and
combined cardiac disease by history, a more severe
motor deficit, and a higher 30-day fatality rate. How-
ever, the groups were similar in their frequency of
lacunar, hemispheral, and basilar syndromes, CT scan
evidence of superficial infarction, degree of worsening
in the hospital, prevalence of stroke prior to their ini-
tial SDB event and prior TIA, race, and history of
diabetes.
Discussion
The specific cause of infarction in a discouragingly
large number of patients continues to elude clinicians
despite efforts to arrive at a diagnosis. Earlier studies
relied heavily upon clinical syndromes and risk factors
in the diagnosis of stroke subtype E5-131. Hyperten-
sion, diabetes, and the absence of valvular heart dis-
ease or atrial arrhythmia in an elderly patient favored
the diagnosis of thrombosis whereas a younger patient
with valvular disease and atrial arrhythmia tended to
carry the diagnosis of embolism. CT scanning enor-
mously improved the differential diagnosis of hemor-
rhage from infarction but did not elucidate the mecha-
nisms of infarction 1141.
More recent studies using modern imaging tech-
niques have allowed the reassessment of diagnoses
based on associated risk factors and have permitted
more rigorous definitions of stroke subtype. One re-
sult of the more frequent use of angiography and
noninvasive Doppler studies has been the finding that
infarcts attributed to atherosclerotic stenosis or occlu-
sion of large arteries occur at a lower frequency than
formerly believed when greater reliance was placed on
the risk factors of age, hypertension, and other signs of
arteriosclerosis E 15, 161.
In the SDB a classification scheme was developed to
group infarcts by causal mechanism as outlined in the
Methods section. Every effort was made to ensure that
these diagnostic subtypes resulted in homogeneous
categories, using conventional laboratory studies such
as CT scanning, angiography where deemed safe and
clinically relevant, noninvasive Doppler studies, and a
388 Annals of Neurology Vol 25 No 4 April 1989
search for cardiac disease and alternative coagulation
states. Diagnostic efforts were sometimes limited by
the degree to which the investigator, when not acting
as the primary care physician, guided the workup. The
basis for the diagnosis was specified in each case. The
use of this rigorous diagnostic scheme resulted in the
diagnosis of infarcts that with a high frequency were
difficult to classify into the traditional subtypes and
that were, by default, classified as IUC.
Some examples of the rare forms of stroke attrib-
uted to meningitis, migraine, lupus anticoagulant, ar-
teritis, and the like may be represented in this material.
However, efforts were made in each patient to estab-
lish the existence of the unusual cause, and all patients
so identified were classified as OTHER. Adding to-
gether all the estimated frequencies with which such
unusual causes present without accompanying evi-
dence of the underlying disease cannot remotely ap-
proach the high frequency of diagnosis of IUC docu-
mented in this study.
The failure to find the cause of an infarct is most
often blamed on the failure to perform the appropriate
tests, the improper timing of otherwise appropriate
tests, and normal or ambiguous findings in appropriate
tests done at the appropriate time. In the current
study, considerable effort at workup was undertaken.
CT scanning was performed in almost 100% of pa-
tients, angiography was performed in 27% and nonin-
vasive studies seeking evidence of extracranial and car-
diac disease were performed in 57% of the patients
with IUC without angiography.
The investigators were all aware of the impact of
timing on the results obtained from many tests, espe-
cially angiography. The decision to undertake angiog-
raphy in a patient with acute stroke was based on many
objective and some subjective variables, not least
among them being the safety and clinical status of the
patient. Our model demonstrates that younger patients
with a superficial infarct on CT, prior TIA, less severe
motor deficit, and presentation without a typical lacu-
nar syndrome were more likely to be selected for an-
giography. There are other variables that might have
been difficult to assess in our model, such as allergies
to contrast; refusal of the patient, family, or referring
physician to consent to angiography; renal disease; or
comorbid conditions raising the risk of the procedure
to intolerable levels. Yet the same implicit algorithm
for angiography that applied to all infarcts also ap-
peared to operate in the patients with IUC. These
patients did not have a disproportionately low fre-
quency of angiography as a reason for failure to arrive
at a cause for the infarct. Despite the use of angiog-
raphy, there remained a large percentage of patients
for whom the infarct mechanism escaped explanation
and who were labeled as having IUC.
The results of angiography can fail to disclose direct
diagnostic information as to the infarct mechanism. Se-
rial angiographic studies demonstrated the disappear-
ance of initial occlusions in only a few days, clarifying
the unstable nature of embolic material and suggesting
that a normal angiogram could easily be found in an
embolic stroke [17, 181. Intracranial stenosis may be
misinterpreted on angiography when, in actuahty, a
recanalized embolus is present 1171. Predicting the
degree of hemodynamic insufficiency of a carotid ste-
nosis or occlusion based on the distribution of collat-
erals may be misleading [20]. There are clearly diag-
nostic limitations to angiography; however, important
inferences can be made based on the relationship of
angiographic findings to the clinical presentation.
One approach to d e d n g with this large cohort of
IUC patients is forcing reclassification into the tradi-
tional categories of atherothrombosis, embolism, or
lacunar infarction. The presentation of a hemispheral
syndrome, a surface infarction by CT, and a corre-
sponding branch occlusion documented by angiog-
raphy or a normal angiogram has long been considered
suggestive of embolism 117-17, 21-27]. There is am-
ple evidence for many occult sources of emboli, the
difficulty proving their existence, and their role in the
first or succeeding ischemic strokes 1281. A proportion
of IUC patients who underwent angiography showed
distal intracranial branch occlusions or normal results
on angiography. Such findings could be inferred to
represent emboli, despite no cardiac source for embo-
lism documented by clinical or laboratory criteria.
Reclassification of such patients and others with lim-
ited cerebral infarction seen on CT scan as embolism
from an undetected source would add the majority of
our cohort of 138 patients with IUC to the EMB
group, making embolism from all sources the largest
subtype of stroke. Rather than reclassifying these cases
as EMB, in which embolism is the implied mechanism,
we suggest their continued separation as IUC, because
they represent a sizable cohort of acute strokes for
which no mechanism has yet been proven. The search
for causes of brain infarction continues to uncover new
potential mechanisms of infarction [29, 30}, perhaps
leading to new forms of therapy.
Analyses on the SDB cohort cannot be generalized
to all patients with stroke. However, essential informa-
tion is provided on patients who are admitted to met-
ropolitan teaching centers and available for randomiza-
tion in clinical trials. The sample size requirements of
current clinical trials have led to lumping of categories
of ischemic stroke 1311, a practice that argues that no
important difference exists between subtypes of isch-
emic stroke. We suggest that examination of prognosis
by infarct subtype will permit better delineation of the
risks of stroke progression and recurrence. Continued
stratification of clinical studies by the diagnostic sub-
types ATH, EMB, LAC, and IUC might even lead to
further refinement or clarification of the group cur-
rently classified as IUC. A better understanding of the
prognosis and mechanism of stroke subtypes will be
extremely useful in the design of efficient clinical stud-
ies of stroke therapy.
This work was supported by the Stroke Data Bank under contracts
N01-NS-2-2302, N01-NS-2-2384, N01-NS-2-2398, NOl-NS-2-
2399, N01-NS-6-2305 from the National Institute of Neurological
and Communicative Disorders and Stroke.
We thank Dr H. J. M. Barnett for his thoughtful suggestions on the
manuscript, Kathryn Chantry for her able assistance in carrying out
the statistical analyses, and Dr Mary A. Foulkes for her guidance in
the Stroke Data Bank project.
This work was presented in part in abstract form at the 11lth An-
nual Meeting of the American Neurological Association, Boston,
MA, October 5-8, 1986.
The SDB manual of operations, which includes the SDB data forms,
is available from the National Technical Information Service, US
Department of Commerce, 5285 Port Royal Road, Springfield, VA
22161 (NTIS Accession No. PB88 101852/AS).
References
1. Foulkes MA, Wolf PA, Price TR, et al. The stroke data bank
design, methods, and baseline characteristics. Stroke 1988;
19:547-554
2. Mohr JP, Nichols FI',Tatemichi TK. Classification and diag-
nosis of stroke. International Angiology 1984;3:431-439
3. Harrell FE. The logistic procedure. In: The SAS Users Group
International supplemental library user's guide. Cary, NC: SAS
Institute, 1983:181-202
4. Gordon T, Kannel WB, Halperin M. Predictability of coronary
artery disease. J Chronic Dis 1979;32:427-440
5. Hyland HH. Thrombosis of intracranial arteries. Arch Neurol
Psychiatry 1933;30:342-3 56
6. Aring CD, Merrin HH. Differential diagnosis between cerebral
hemorrhage and cerebral thrombosis. Arch Intern Med 1935;
56:435-456
7. Eisenberg H, Morrison JT, Sullivan P, et al. Cerebrovascular
accidents: incidence and survival rates in a defined population,
Middlesex County, Connecticut. JAMA 1964;189:883-888
8. Lhermitte F, Gautier JC, Derouesne C, Guiraud B. Ischemic
accidents in the middle cerebral artery territory-a study of the
causes in 122 cases. Arch Neurol 1968;19:248-256
9. Alter M, Christoferson L, Resch J, et al. Cerebrovascular dis-
ease: frequency and population selectivity in an upper Midwest-
ern community. Stroke 1970;1:454-465
10. Kannel WB, Dawber TR, Cohen MS, et al. Vascular disease of
the brain-epidemiologic aspects: the Framingham study. Am J
Public Health 1965;55:1355-1366
11. Whisnant JP, Fitzgibbons JP, Kurland LT, et al. Natural history
of stroke in Rochester, Minnesota, 1945 through 1954. Stroke
197 1;2:11-22
12. Matsumoto N, Whisnant JP, Kurland LT, et al. Natural history
of stroke in Rochester, Minnesota, 1955 through 1969: an ex-
tension of previous study, 1945 through 1954. Stroke 1973;
4:20-29
13. Mohr JP, Caplan LR,MelskiJW, et al. The Harvard cooperative
stroke registry: a prospective registry. Neurology 1978;28:754-
762
Sacco et al: Infarcts of Undetermined Cause 389
14. Drury I, Whisnant JP, Garraway WM. Further observations on
intracerebral hemorrhage in Rochester, Minnesota, with CT
confirmation. Neurology 1983;33(suppl2):84
15. Gross CR, Kase CS, Mohr JP, et al. Stroke in South Alabama:
incidence and diagnostic features. Stroke 1984;15:249-255
16. Kunitz SC, Gross CR, Heyman A, et al. The pilot stroke data
bank: definition, design, and data. Stroke 1984;15:740-746
17. Dalal PM, Shah PM, Aiyar RR. Arteriographic study of cerebral
embolism. Lancet 1965;2:358-361
18. Liebeskind A, Chinichian A, Schechter MM. The moving em-
bolus seen during serial cerebral angiography. Stroke 1971;
2~440-443
19. Irino T, Tandea M, Minami T. Angiographic manifestations in
postrecanalized cerebral infarction. Neurology 1977;27:47 1-
475
20. Powers WJ, Press GA, Grubb RL, et al. The effect of hemody-
namically significant carotid artery disease on the hemodynamic
status of the cerebral circulation. Ann Intern Med 1987;27:106
21. Samuel KC. Atherosclerosis and occlusion of the internal
carotid artery. J Pathol Bacteriol 1956;71:391-401
22. Castaigne P, Lhermitte F, Gautier JC, et al. Internal carotid
artery occlusion: a study of 61 instances in 50 patients with post-
mortem data. Brain 1970;93:231-258
23. Fisher CM, Gore I, Okabe N, White PD. Atherosclerosis of the
390 Annals of Neurology Vol 25 No 4 April 1989
carotid and vertebral arteries-extracranial and intracranial. J
Neuropathol Exp Neurol 1965;24:455-476
24. Torvik A, Jorgensen L. Thrombotic and embolic occlusions of
the carotid arteries in an autopsy material: Part 2. Cerebral
lesions and clinical course. J Neurol Sci 1966;3:410-432
25. Ring BA. Diagnosis of embolic occlusions of smaller branches
of the intracerebral arteries. Am J Roentgen01 1966;97:575-
582
26. Davis DO, Rumbaugh CL, Gilson JM. Angiographic diagnosis
of small vessel cerebral emboli. Acta Radio1 [Diagn) (Stockh)
1969;9:264-27 1
27. Bladin PF. A radiologic and pathologic study of embolism of
the internal carotid-middle cerebral arterial axis. Radiology
1964;82:614-624
28. Mohr JP. Infarct of unclear cause. In: Furlan AJ, ed. The heart
and stroke-exploring mutual cerebrovascular and cardiovascu-
lar issues. New York: Springer-Verlag, 1987:lOl-116
29. Sacco RL, Owen J, Tatemichi TK,Mohr JP. Protein S deficiency
and intracranial vascular occlusion. Ann Neurol 1987;22:115
30. Levine SR, Kim S, Deegan MJ, Welch KMA. Ischemic stroke
associated with anti-cardiolipin antibodies. Stroke 1987;18:
1101-1106
31. The ESPS Group. The European Stroke Prevention Study
(ESPS)-principal end-points. Lancet 1987;2:135 1-1354