Sie sind auf Seite 1von 5

Stroke Classification

A Personal View
Louis R. Caplan, MD

B ecause all history is to some extent personal, and because


I have been active throughout attempts at classification, I
was asked to provide a personal account of stroke subtyping and
clinical information into a computer. The computer used this
data to identify an acid– base abnormality, eg, metabolic
acidosis, and to calculate the content and amount of infusate
stroke registries bringing the topic up to date. needed to correct the imbalance. After using the program,
clinicians learned the data needed and, by reading references
Early Studies appended to the program, learned how the computer made
Clinicopathological studies during the first half of the 20th diagnoses and calculations. Doctors could then master the
century focused on clinical signs in patients who died after process without using the computer. The program used a
stroke. Brain hemorrhages and infarcts were recognized at large PDP-11 computer housed in a temperature-controlled
necropsy, but could they be separated during life? In 1935, room to ensure that the computer would not break and lose all
Downloaded from http://stroke.ahajournals.org/ by guest on March 5, 2018

Aring and Merritt1 analyzed 245 stroke patients studied data. Bleich asked me to construct a computer-based program
clinically and at necropsy at the Boston City Hospital. They that could diagnose all neurological disease similar to the
reported demographic, epidemiological, historical, and clini- program for renal and acid– base abnormalities. I found the
cal data to determine which features differentiated thrombosis task of confronting all neurological conditions to be much too
from hemorrhage. Hemorrhages (intracerebral or subarach- formidable, but after cajoling I agreed to construct a program
noid) were found in 15% of patients and 82% had ischemic aimed only at stroke diagnosis that considered only stroke
infarcts called “thrombotic;” however, only 3% were consid-
etiology subtyping and did not consider brain localization.
ered to have cardiogenic embolic brain infarcts. The main
At that time, I was completely computer illiterate. A very
features favoring hemorrhage were headache, vomiting, im-
bright medical student and Massachusetts Institutes of Technol-
paired consciousness, progression of the clinical neurological
ogy graduate, Robert Goldstein, was assigned to work with me.
deficit, bloody spinal fluid, and increased spinal fluid pres-
Goldstein and Bleich explored with me potential strategies that
sure.1 This series was biased toward patients with large fatal
could be used to construct a prospective computer-based diag-
hemorrhages and infarcts. Later, Dalsgaard-Nielsen2 in Scan-
nostic program. Branching logic and pattern-matching proved
dinavia compiled a series of 1 000 stroke patients, and
unsuitable because no large compilation of patients to pattern-
clinicians at the Mayo Clinic analyzed series of patients seen
match was available, and there were few branches that were
during 1945 through 19543 and from 1955 through 1969.4
absolutely indicative of 1 stroke subtype. The only feasible
These studies were based on retrospective chart reviews
and all preceded CT. Infarcts were ⬎4-times more common strategy was using a Bayesian analysis. Bayes theorem involved
than hemorrhages. Infarcts were classified as embolic only if calculating the likelihood of a particular diagnosis based on the
the patients had rheumatic heart disease or recent myocardial frequency of a condition in the population and the frequency of
infarction. Rates of brain embolism using these criteria particular findings in that condition.
ranged from 3% to 8%. Nonembolic brain infarcts were I buried myself in the library but could not find relevant data
assumed to be “thrombotic” and related to occlusion of brain from the literature. Carotid artery disease, cardiac lesions other
supplying large arteries. Carotid and vertebral artery disease than rheumatic and acute myocardial infarction, and lacunar
in the neck and lacunar infarction were not diagnoses in- infarcts were recently described and there was no quantitative
cluded in any of these early studies. data about the various clinical findings in these conditions. Brain
infarction was known to be present in ⬇80% of strokes, and
The Harvard Stroke Registry5– 8 subarachnoid hemorrhage and intracerebral hemorrhage each
In 1971, Howard Bleich, a nephrologist at the Beth Israel accounted for ⬇10% of strokes. So, in an emergency depart-
Hospital in Boston, Massachusetts, published a computer- ment, the diagnosis of brain infarction would be correct 4 times
based diagnostic program guiding diagnosis and treatment of out of 5, but if one diagnosed intracerebral hemorrhage in all
acid– base abnormalities encountered in hospitalized pa- patients, the accuracy would be only 10%. The frequency of
tients.6 This program required clinicians to enter patients’ individual findings, various risk factors, headache preceding

Received June 26, 2010; accepted October 12, 2010.


From the Beth Israel Deaconess Medical Center, Department of Neurology, Boston, Mass.
Correspondence to Louis R. Caplan, MD, Beth Israel Deaconess Medical Center, Department of Neurology, Palmer 127, West Campus, 330 Brookline
Avenue, Boston, MA 02215-5400. E-mail lcaplan@bidmc.harvard.edu
(Stroke. 2011;42[suppl 1]:S3-S6.)
© 2010 American Heart Association, Inc.
Stroke is available at http://stroke.ahajournals.org DOI: 10.1161/STROKEAHA.110.594630

S3
S4 Stroke January 2011

stroke and at onset, TIA preceding stroke, progression of the and racial diversity was important. The Stroke Data Bank was
clinical deficit, vomiting, loss of consciousness, and so on, were created and funded by the National Institute of Neurological
not available from the literature. The next step was to ask Disease and Stroke. The institutions and principle investigators
experienced clinicians to estimate the various frequencies. Drs were J.P. Mohr at Columbia University, Tom Price at the
Miller Fisher, Raymond Adams, Richard Tyler, J.P. Mohr, and University of Maryland, Phil Wolfe at Boston University, and I
I made frequency guesses. To our surprise, the various estimates at Michael Reese Hospital in Chicago. During the collection
were all over the ballpark. The estimate of some single findings period, Dan Hier became the principle investigator in Chicago.
varied from 80% to 10%. A Massachusetts Institutes of Tech- During the first year of the grant, the principle investigators
nology graduate averaged the estimates, and Goldstein, Bleich, and their assistants (including Ralph Sacco and Stan Tuhrim)
and I constructed a computer-based diagnostic program that we met often in Bethesda with National Institute of Neurological
used to separate subarachnoid hemorrhage, intracerebral hemor- Disease and Stroke designates to generate the terms, data
rhage, brain embolism, large artery-related brain infarction, and entry items, main queries to be answered, and statistical
lacunar infarction using clinical data.7 methods. During 1983 to 1986, 1 805 patients (54% black)
The main outcome of this attempt at computer diagnosis were prospectively examined and the results were published
was to make me aware that there were little data available during the late 1980s and early 1990s.9 –15
about the frequency of various clinical findings in the various In the interval between patient enrollment in the Harvard Stroke
subtypes of stroke. So, I decided that we needed to prospec- Registry and the Stroke Data Bank, technology had changed.
tively collect that data. The Harvard Cooperative Stroke Nearly all patients (97%) in the Stroke Data Bank had brain CT
Downloaded from http://stroke.ahajournals.org/ by guest on March 5, 2018

registry was born. I asked Dr J.P. Mohr, who was the chief of scans and many had echocardiography. Few had cerebral an-
the Stroke Service at the Massachusetts General Hospital, to giography; MRI, MRA, and CTA were not available. Cardiac
join me in collecting data. Mohr and I met with Bleich, conditions were characterized as high risk or medium risk.
Warner Slack, and John Melski, then a computer fellow at the Cortical or cortical/subcortical infarcts in the Stroke Data Bank
Beth Israel Hospital, on a weekly basis to decide on classi- were labeled infarct of undetermined cause, unless there was a
fication criteria, entry items, and terms. high-risk cardiac source. If a large neck artery was occluded and
In the Harvard Stroke Registry, the diagnosis of embolism studies showed that a distal branch was also occluded, then the
was defined as blockage of a distal vessel by material diagnostic category was tandem arterial pathology. Only 6% of
generated at a distance. The evidence used related mostly to patients qualified for large artery stenosis or occlusion. Reacting
the recipient artery. The source could be cardiac, aortic, to the criticism that the Harvard Stroke Registry had overdiag-
intra-arterial, or unknown. At that time (early 1970s), the nosed embolism, the Stroke Data Bank used different criteria
major diagnostic tool was cerebral angiography and echocar- that emphasized the embolic source rather than the recipient
diography was not available. artery. The frequencies were: infarcts, 70% (large artery 6%,
The frequencies of diagnosis among the 694 patients in the tandem pathology 4%, lacunes 19%, cardiac source 14%,
Harvard Stroke Registry were: thrombosis, 53% (34% large undetermined cause 28%); hemorrhages, 27% (subarachnoid
artery, 19% lacunar); embolism, 31%; intracerebral hemor- hemorrhage and intracerebral hemorrhage 13% each); and other,
rhage, 10%; and subarachnoid hemorrhage, 6%.5 Some cases 3%. The sum of those diagnosed as cardiac source embolism and
were classified as unsettled cause and some as unusual cause, infarct unknown cause (nonlacunar infarcts without large artery
and these were not included in the frequencies cited. Also occlusions or tandem arterial pathology) totaled 42%. These
noted were the basis of the diagnosis (clinical, angiography, patients were classified as embolism in the Harvard Stroke
necropsy) and confidence that the diagnosis was correct, high Registry.
or low. The basis for the diagnoses in approximately half of The presence of CT scan data now allowed study of the
the patients was only clinical; 45% had cerebral angiography, distribution of infarcts. The infarct lesions were sketched on
only 3% had CT scans, and 4% had necropsies. standard grid template grids that covered 10 CT sections.
The Harvard Stroke Registry was the first prospective Correlations could be made with neurological signs and
published database on any medical condition. The results stroke mechanisms.11–15
were submitted in a manuscript to the New England Journal
of Medicine. The rejection letter commented that computers New England Medical Center Posterior
would never take hold in medicine. The registry ended when Circulation Registry17–25
Dr Mohr and I both left Boston. Knowledge that the advent of After moving to Boston in 1984, I began a registry limited to
CT scanning would forever change diagnosis led us to begin patients studied at the New England Medical Center (NEMC)
new data collections based more on brain imaging. who had ischemic disease of the vertebral basilar arterial system.
The aims and rationale of the NEMC Posterior Circulation
The Stroke Data Bank Registry (NEMC-PCR) were quite different than those in the
Although the National Institute of Neurological Disease and Harvard Stroke Registry and Stroke Data Bank. The focus was
Stroke had rejected the original Harvard Stroke Registry appli- narrowed to ischemia (rather than all strokes) and to the posterior
cation for funding, they became impressed with the idea of a circulation rather than the entire brain. By the end of the 1980s,
government-sponsored prospective stroke registry and issued a CT scans became generally available and brain hemorrhages
request for proposals for a multicenter Stroke Registry. The were readily diagnosed by imaging. Attention turned to differ-
patients studied in the Harvard Stroke registry were from one entiating the various subtypes of brain ischemia. By then, the
institution and were predominantly white. More geographical TOAST trial investigators had published criteria for diagnosing
Caplan Stroke Classification S5

Figure 2. Brain locations vs stroke mechanisms in the New


Figure 1. Brain locations within the proximal territory in the New England Medical Center Posterior Circulation Registry. IA,
England Medical Center Posterior Circulation Registry. PICA, intra-arterial.
posterior inferior cerebellar artery.
Similarly, in a patient with atrial fibrillation with sudden onset of
Downloaded from http://stroke.ahajournals.org/ by guest on March 5, 2018

subtypes of ischemic stroke.16 I had been the chair of the ad hoc an occipital infarct and also 50% stenosis of the left vertebral
committee that reviewed the trial for the National Institute of artery, the most likely mechanism would be cardiac-source
Health and I was well aware of the criteria used. embolism, but the large artery lesion would be included within
Patients were entered into the NEMC-PCR, a prospective the range of large artery disease. In this way, all potential
computerized registry between 1988 and 1996. By then, MRI mechanisms would be tabulated.
had become available and could show the brain stem and In the NEMC-PCR, brain lesions were categorized as involv-
cerebellum better than CT. MRA and transcranial Doppler ing proximal, middle, and distal intracranial posterior circulation
were helpful in studying occlusive vascular lesions. The territories. Proximal posterior circulation territory included in-
NEMC-PCR contained 407 prospectively collected, thor- tracranial vertebral arteries-supplied regions, the medulla and
oughly studied, predominantly white and Asian patients. posterior inferior cerebellar artery territory of the cerebellum.
Vascular studies, such as ultrasound, MRA, or catheter The middle intracranial posterior circulation territory included
angiography, were performed in almost all patients and all brain regions supplied by the basilar artery up to its superior
had brain imaging (predominantly MRI). cerebellar artery branches, the pons and anterior inferior cere-
By the 1990s, it was possible to analyze the presence of bellar artery-supplied cerebellum. The distal territory included
vascular lesions, brain locations, stroke mechanisms, and neu- regions supplied by the rostral basilar artery, superior cerebellar
rological symptoms and signs and to relate risk factors to artery, posterior cerebral artery, and their penetrating artery
vascular lesions and stroke mechanisms, and to relate brain branches—midbrain, thalamus, superior cerebellar artery cere-
locations and symptoms and signs to vascular lesions and stroke bellum, and posterior cerebral artery territories.
mechanisms.17–19 Outcomes also could be analyzed in relation to Brain imaging (CT and/or MRI) was performed on all
brain location, vascular lesions, and stroke mechanisms.20 Be- patients (⬎80% had MRI). Vascular imaging was also
cause designation of stroke mechanism was controversial in the performed in all patients (80% had contrast catheter angiog-
Harvard Stroke Registry and the Stroke Data Bank, we chose a raphy). Ultrasound was widely used (⬎80% of patients had
different strategy in the NEMC-PCR. After review and re- transcranial Doppler). Echocardiography and heart rhythm
review of each patient’s studies, Drs Caplan, Pessin, DeWitt, monitoring were performed when clinically indicated.
Tapia, and the Stroke fellows present at that time decided on the Our major aims in the registry were to clarify outcomes and
most likely stroke mechanism but also recognized any other the frequency of various stroke mechanisms and vascular
potential stroke mechanisms. Tabulation of stroke mechanisms lesions, and to understand the relationship of these mecha-
consisted of 2 columns, one column that contained the consensus nisms and vascular lesions to the topography of brain infarcts.
selection of the most likely stroke mechanism and a second Figure 1 shows the distribution of infarcts within the proxi-
column of ranges of potential mechanisms. The lowest percent- mal territory, and Figure 2 shows the distribution of lesions
age would be if the minimum number with that attribute was according to stroke mechanisms.
responsible, and the highest frequency in the range would be if
all were culpable, for example, if a patient with a bifid stenotic A-S-C-O (Phenotypic) Stroke Classification26,27
aortic valve had occlusion of the right intracranial vertebral During the past decade, I and a group of international stroke
artery and had experienced multiple TIA referable to territory leaders (Amarenco [France], Donnan [Australia], Bogous-
supplied by that artery, followed by a lateral medullary infarct. slavsky [Switzerland], and Hennerici [Germany]) met and de-
The single most likely mechanism would be large artery occlu- vised a new type of Ischemic Stroke Registry. We called it the
sive disease. However, because there was a potential cardiac A-S-C-O (Phenotypic) Stroke Classification, in which A indi-
source (although unlikely to be culpable in this case), it would be cated atherosclerosis, S indicated small vessel disease, C indi-
included in the cardiac embolism source second column range. cated cardiac disease, and O indicated other. The registry, like
S6 Stroke January 2011

Table. Grades of Likelihood Within Each Phenotype a cardiac source of embolism in the NINCDS Stroke Data Bank: his-
torical features. Neurology. 1990;40:281–284.
Grade 1: Definitely a potential cause of the index stroke 11. Timsit SG, Sacco RL, Mohr JP, Foulkes MA, Tatemichi TK, Wolf PA,
Grade 2: Causality uncertain Price TR, Hier DB. Brain infarction severity differs according to cardiac
or arterial embolic source. Neurology. 1993;43:728 –733.
Grade 3: Unlikely a direct cause of the index stroke but disease is present 12. Mohr JP, Foulkes MA, Polis AT, Hier DB, Kase CS, Price TR, Tatemichi
Grade 0: No disease TK, Wolf PA. Infarct topography and hemiparesis profiles with cerebral
convexity infarction: the Stroke Data Bank. J Neurol Neurosurg Psychiatry.
Grade 9: Cannot grade because no tests performed 1993;56:344–351.
Levels of evidence according to diagnostic testing 13. Kittner SJ, Sharkness CM, Sloan MA, Price TR, Dambrosia JM, Tuhrim
S, Wolf PA, Mohr JP, Hier DB. Features on initial computed tomography
A: Direct demonstration by gold standard diagnostic tests or criteria scan of infarcts with a cardiac source of embolism in the NINDS Stroke
B: Indirect evidence or less sensitive or specific tests or criteria Data Bank. Stroke. 1992;23:1748 –1751.
14. Kittner SJ, Sharkness CM, Sloan MA, Price TR, Dambrosia JM, Tuhrim
C: Weak evidence in the absence of specific tests or criteria S, Wolf PA, Mohr JP, Hier DB, Caplan LR. Infarcts with a cardiac source
of embolism in the NINDS Stroke Data Bank: neurologic examination.
Neurology. 1992;42:299 –302.
the NEMC-PCR, was limited to brain ischemia. Data entry made 15. Caplan LR, Kelly M, Kase CS, Hier DB, White JL, Tatemichi T, Mohr J,
information useful for a variety of purposes, such as describing Price T, Wolf P. Infarcts of the inferior division of the right middle
cerebral artery: mirror image of Wernicke’s aphasia. Neurology. 1986;
patient characteristics in therapeutic trials, grouping patients in
36:1015–1020.
epidemiological studies, phenotyping in genetic studies, and 16. Low molecular weight heparinoid, ORG 10172 (danaparoid), and
classifying patients for therapeutic decision-making in daily outcome after acute ischemic stroke: a randomized controlled trial. The
Publications Committee for the Trial of ORG 10172 in Acute Stroke
Downloaded from http://stroke.ahajournals.org/ by guest on March 5, 2018

practice. We defined each subtype clearly. The system was


Treatment (TOAST) Investigators. JAMA 1998;279:1265–1272.
designed to include the potential likelihood of the presence of 17. Caplan LR, Wityk RJ, Glass TA, Tapia J, Pazdera L, Chang HM, Teal P,
each stroke subtype and the probability that it caused the Dashe JF, Chaves CJ, Breen JC, Vemmos K, Amarenco P, Tettenborn B,
index event. Definitions and the levels of evidence were Leary M, Estol C, Dewitt LD, Pessin MS. New England Medical Center
Posterior Circulation registry. Ann Neurol. 2004;56:389 –398.
agreed on. The likelihood of the presence of each phenotype 18. Caplan L, Chung, CS, Wityk R, Glass T, Tapia J, Pazdera L, Chang HM,
was assigned grades and levels of evidence (Table). Dashe J, Chaves C, Vemmos K, Leary M, Dewitt L, Pessin M. New
England medical center posterior circulation stroke registry: I. Methods,
data base, distribution of brain lesions, stroke mechanisms, and outcomes.
Disclosures J Clin Neurol. 2005;1:14 –30.
None. 19. Caplan L, Wityk R, Pazdera L, Chang HM, Pessin M, Dewitt L. New
England Medical Center posterior circulation stroke registry II. Vascular
References lesions. J Clin Neurol. 2005;1:31– 49.
1. Aring CD, Merritt HH. Differential diagnosis between cerebral hemor- 20. Glass TA, Hennessey PM, Pazdera L, Chang HM, Wityk RJ, DeWitt LD,
rhage and cerebral thrombosis. Arch Intern Med. 1935;56:435– 456. Pessin MS, Caplan LR. Outcome at 30 days in the New England Medical
2. Dalsgaard-Nielsen T. Survey of 1000 cases of apoplexia cerebri. Acta Center Posterior Circulation Registry. Arch Neurol. 2002;59:369 –376.
Psychiatr Neurol Scand. 1955;30:169 –185. 21. Wityk RJ, Chang HM, Rosengart A, Han WC, DeWitt LD, Pessin MS,
Caplan LR. Proximal extracranial vertebral artery disease in the New
3. Whisnant JP, Fitzgibbons JP, Kurland LT, Sayre GP. Natural history of
England Medical Center Posterior Circulation Registry. Arch Neurol.
stroke in Rochester, Minnesota, 1945 through 1954. Stroke. 1971;2:11–22.
1998;55:470 – 478.
4. Matsumoto N, Whisnant JP, Kurland LT, Okazaki H. Natural history of
22. Müller-Küppers M, Graf KJ, Pessin MS DeWitt LD, Caplan LR. Intra-
stroke in Rochester, Minnesota, 1955 through 1969: an extension of a
cranial vertebral artery disease in the New England Medical Center
previous study, 1945 through 1954. Stroke. 1973;4:20 –29.
Posterior Circulation Registry. Eur Neurol. 1997;37:146 –156.
5. Mohr JP, Caplan LR, Melski JW, Goldstein RJ, Duncan GW, Kistler JP, 23. Shin HK, Yoo KM, Chang HM, Caplan LR. Bilateral intracranial ver-
Pessin MS, Bleich HL. The Harvard Cooperative Stroke Registry: a tebral artery disease in the New England Medical Center, Posterior
prospective registry. Neurology. 1978;28:754 –762. Circulation Registry. Arch Neurol. 1999;56:1353–1358.
6. Bleich HL. The computer as a consultant. N Engl J Med. 1971;284: 24. Voetsch B, DeWitt LD, Pessin MS, Caplan LR. Basilar artery occlusive
141–147. disease in the New England Medical Center Posterior Circulation
7. Goldstein RJ, Caplan LR, Bleich HL, Melski JW, Mohr JP. Computer Registry. Arch Neurol. 2004;61:496 –504.
stroke registry and diagnosis program [Abstract]. Neurology. 1976; 25. Yamamoto Y, Georgiadis AL, Chang HM, Caplan LR. Posterior cerebral
25:356. artery territory infarcts in the New England Medical Center Posterior
8. Melski JW, Caplan LR, Mohr JP, Geer DE, Bleich HL. Modeling the Circulation Registry. Arch Neurol. 1999;56:824 – 832.
diagnosis of stroke at two hospitals. MD Computing. 1989;6:157–163. 26. Amarenco P, Bogousslavsky J, Caplan LR, Donnan GA, Hennerici MG.
9. Foulkes MA, Wolf PA, Price TR, Mohr JP, Hier DB. The Stroke Data Bank: New approach to stroke subtyping: the A-S-C-O (phenotypic) classifi-
design, methods, and baseline characteristics. Stroke. 1988;19:547–554. cation of stroke. Cerebrovasc Dis. 2009;27:502–508.
10. Kittner SJ, Sharkness CM, Price TR, Plotnick GD, Dambrosia JM, 27. Amarenco P, Bogousslavsky J, Caplan LR, Donnan GA, Hennerici MG.
Tuhrim S, Wolf PA, Mohr JP, Hier DB, Kase CS, Tuhrim S. Infarcts with Classification of stroke subtypes. Cerebrovasc Dis. 2009;27:493–501.
Stroke Classification: A Personal View
Louis R. Caplan

Stroke. 2011;42:S3-S6; originally published online December 16, 2010;


Downloaded from http://stroke.ahajournals.org/ by guest on March 5, 2018

doi: 10.1161/STROKEAHA.110.594630
Stroke is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231
Copyright © 2010 American Heart Association, Inc. All rights reserved.
Print ISSN: 0039-2499. Online ISSN: 1524-4628

The online version of this article, along with updated information and services, is located on the
World Wide Web at:
http://stroke.ahajournals.org/content/42/1_suppl_1/S3

Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally published
in Stroke can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial Office.
Once the online version of the published article for which permission is being requested is located, click
Request Permissions in the middle column of the Web page under Services. Further information about this
process is available in the Permissions and Rights Question and Answer document.

Reprints: Information about reprints can be found online at:


http://www.lww.com/reprints

Subscriptions: Information about subscribing to Stroke is online at:


http://stroke.ahajournals.org//subscriptions/

Das könnte Ihnen auch gefallen