Decisions Investigation stroke: strategy?

Postgrad Med J (1991) 67, 259 - 270 © The Fellowship of Postgraduate Medicine, 1991

Difficult Decisions

Investigation of acute stroke: what is the most effectivestrategy?

D.W. Dunbabin and P.A.G. Sandercock

Department of Clinical Neurosciences, Western General Hospital, Edinburgh EH4 2XU, UK

Summary: Techniques of investigation of acute stroke syndromes have progressed rapidly in recentyears, outpacing developments in effective stroke treatment. The clinician is thus faced with a variety oftests, each with different cost implications and each altering management to a greater or lesser extent. Thisreview will concentrate on the basic tests which should be performed for all strokes (full blood count, ESR,biochemical screen, blood glucose, cholesterol, syphilis serology, chest X-ray and electrocardiogram).Additional tests may be required in selected cases: CT scan to diagnose 'non-stroke' lesions, to excludecerebral haemorrhage if anti-haemostatic therapy is planned, and to detect strokes which may requireemergency intervention (such as cerebellar stroke with hydrocephalus); echocardiography to detectcardiac sources of emboli; and in a few cases lumbar puncture and specialized haematological tests. Othertests, which are currently research tools, may be suitable for widespread use in the future including NMR,SPECT and PET scanning.

Introduction

The problem of stroke

In the next year about 100,000 people in the UnitedKingdom will have a definite first stroke and manyothers will have symptoms suggestive of stroke. Ofthe definite strokes about 20,000 will die and about30,000 survivors will be seriously disabled.' Afterrecovery from stroke the survivors will be at risk ofrecurrence and other vascular events such asmyocardial infarction and vascular death. Theeconomic burden of stroke is large, approximately5% of the total National Health Service budget.2There will be additional direct costs to the SocialServices and indirect costs to the community fromthe lost income of unpaid carers.About 85% of strokes are due to cerebral

infarction (caused by occlusion of extracranial orintracranial cerebral vessels), 10% are due toprimary intracerebral haemorrhage and 5% aredue to subarachnoid haemorrhage.3 Stroke is aheterogeneous disease with a wide variety ofclinical presentations, each associated withdiffering prognosis. The clinical management ofpatients in the acute phase of stroke remainscontroversial. Imaging and cerebral function tech-niques have advanced enormously, yet the mostappropriate scheme of investigation and the most

effective treatment ofa patient with suspected acutestroke is not clear, a contrast to the investigationand management of other occlusive vasculardiseases such as myocardial infarction. This reviewwill concentrate on strategies for the investigationof all forms of suspected stroke with the exceptionof the clinically distinct syndrome of subarachnoidhaemorrhage, which has recently been reviewedvery well by Vermeulen.4 We will also briefly reviewthe general principles of the assessment of diagnos-tic tests and then the value of specific tests in theassessment of suspected stroke.

Why investigate a stroke?

'Errors injudgment must occur in the practice ofanart which consists largely in balancing proba-bilities.' Sir William Osler.

The reasons for investigating an individual patientpresenting with suspected acute stroke are varied.Decision analysis with formal modelling may pro-vide a better way of selecting appropriate tests, butis still a relatively crude indicator of the mostcost-effective form of investigation. Algorithmflowcharts can provide a rigid protocol to follow,but may need to be altered as technology advances,or to suit different institutions (with differingpatterns of disease presenting and differing inves-tigational facilities), or to suit the views of different

Correspondence: P.A.G. Sandercock, D.M., M.R.C.P.Received: 13 September 1990

copyright. on N

ovember 8, 2021 by guest. P

rotected byhttp://pm

j.bmj.com

/P

ostgrad Med J: first published as 10.1136/pgm

j.67.785.259 on 1 March 1991. D

ownloaded from

http://pmj.bmj.com/

260 D.W. DUNBABIN & P.A.G. SANDERCOCK

physicians (depending on their view of what con-stitutes the best treatment). Nonetheless, a system-atic approach should ensure that the investigationof patients with stroke becomes more rational,more cost-effective and less risky.

General aims ofinvestigation

(a) Confirm the diagnosis of stroke The maindifferential diagnoses for stroke are listed in TableI. The majority of these can be excluded by carefulattention to establishing the history. If there is aclear history of sudden onset, non-stroke lesionsare very unlikely. If there is doubt about speed ofonset or progression or if the patient is confused,dysphasic or comatose, a telephone call (cost, a fewpence) to talk to a witness or member of the closefamily may be the most cost-effective investigationand is certainly easier to obtain than a computedtomographic (CT) scan (cost £50-80). Surveys ofadmissions with suspected stroke to stroke unitsand neurological wards reveal a 5-13% incidenceof 'non-stroke' lesions, largely unrecognizedseizures or confusional states.5-' In a community-based sample of patients the proportion of 'non-stroke' lesions will be lower 1.5%,3 reflectingadmission bias; a patient with an atypical clinicalpicture associated with a 'non-stroke' lesion ismore likely to be admitted to hospital.

(b) Clarify the pathology The distinction be-tween cerebral infarction and primary intracere-bral haemorrhage may be important ifanti-haemo-static treatment is contemplated. A scoring systembased on clinical features enables the calculation ofa probability ratio for cerebral haemorrhage,7 butthis may incorrectly classify some small deephaemorrhages as infarcts. The mechanism of in-farction may be identified; for exampledifferentiating between arterial or venous occlusion

Table I Nonvascular lesions which can present with theclinical syndrome of stroke

intracranial tumour (primary or metastatic)subdural haematomapost-ictal paralysis (Todd's paresis)intracranial infection (encephalitis, meningitis orabscess)

hypoglycaemiahyperglycaemia (especially nonketotic)electrolyte disturbance (low serum sodium, phosphate

etc.)multiple sclerosis (rare)occult head injuryperipheral nerve lesionhysteriamigraine

and the type of occlusion may sometimes bedetermined (whether there is in situ thrombosis orembolism from the heart or from atheromatousarterial plaques).

(c) To discover the cause There are manypotential causes ofcerebral infarction, but the mostcommon is arterial occlusion, related to embolicand thrombotic consequences of atheroscleroticdisease of arteries to the brain. A small proportionof strokes (Table II) are related to non-athero-matous, non-embolic causes and are importantbecause they are often amenable to treatment.

(d) As a guide to prognosis Some techniquesprovide information which predicts prognosis, butclinical data such as age, level of consciousness, theseverity of neurological defect, perceptual defectand the development of incontinence are powerfulpredictors of outcome;-'2 studies are now neededto see whether new imaging or laboratory tech-nology, if added to current clinical prognosticscoring systems, improves prognostic accuracy.

(e) As a guide to treatment

(i) Acute phase of stroke. For the majority ofstrokes there is no currently available, widelypracticable treatment which has been demon-strated to improve survival or reduce the functionalhandicap after stroke. Many clinicians feel thatthere is little point in extensive investigation ofpatients soon after stroke onset if no effectivetreatment is available. This rather nihilistic view ofstroke ignores the small proportion ofpatients whohave a non-stroke lesion or a treatable type ofstroke (Tables I and II). If effective treatments forstroke become available it will be important toidentify those who may benefit. For example, ifearly thrombolysis and arterial recanalization isshown to improve outcome it will become impor-tant to identify patients with a cerebral infarction,persisting occlusion of a major cerebral artery, andwith potentially viable ischaemic tissue beforetreatment is started.(ii) Long-term secondary prevention. If long-termtreatment with antiplatelet or anticoagulants isplanned, or if the patient is at high risk ofdeep veinthrombosis (and may require anticoagulation) it isnecessary to exclude primary intracerebral haem-orrhage before starting treatment.

(f) To make or to save money Investigatingpatients purely for profit is not a practice whichmost clinicians would readily admit to, but may bea powerful determinant ofmedical practice in someparts of the world. More relevant to the NationalHealth Service, an investigation may, by avoiding adisabling mis-diagnosis or by identifying appropri-

copyright. on N



j.bmj.com

/P


j.67.785.259 on 1 March 1991. D

ownloaded from

http://pmj.bmj.com/

INVESTIGATION OF ACUTE STROKE 261

Table II Causes of stroke which need specific treatmentor need to be identified for other reasons

CEREBRAL INFARCTIONinfective arteritis- local sepsis, syphilisinflammatory arteritis - giant-cell, systemic lupus

erythematosus, Moya-Moya, fibromuscular dysplasia,polyarteritis nodosa, granulomatous angiitis

cardiogenic embolism - rheumatic heart disease,myxomatous mitral valve disease, endocarditis, muralthrombus, paradoxical embolus, atrial myxoma,cardiac arrhythmia

arterial dissection - traumatic, spontaneous, Marfan's,homocystinuria

coagulopathy - sex steroids, protein C & S deficiency,hyperviscosity syndromes, polycythaemia,anticardiolipin antibodies, sickle-cell disease

drugs - cocaine, amphetaminessystemic hypotension - antihypertensive drugs, surgery,

blood loss, bradycardiadecompression sicknessgas orfat emboluscerebellar infarction with obstructive hydrocephaluscerebral venous thrombosisPRIMARY INTRACEREBRAL HAEMORRHAGEcoagulopathy - iatrogenic anticoagulant overdose,

disseminated intravascular coagulation, haemophiliacongenital vascular malformationsimpairedplatelet number orfunction - leukaemia,

idiopathic thrombocytopaenic purpura, aplasiadrugs - cocaineamyloid degenerationcerebellar haemorrhage with obstructive hydrocephaluscerebral venous thrombosis

influenced by the prevalence of disease in thepopulation assessed and this must be taken intoaccount when interpreting any test result. Mosttests perform best when the prevalence of disease(pre-test probability) is about 50%; a positive ornegative test result can then contribute significantlyto the diagnosis by substantially increasing ordecreasing the post-test probability of disease. Forexample, when the pre-test probability is low (say10%) or high (say 90%), then the result of thediagnostic test will change those probabilities onlyslightly and add little to the clinical diagnosis.13 Anillustration of the ability of Doppler ultrasoundstudies to detect significant carotid stenosis indifferent types of patient with differing pre-testprobabilities of disease is given in Table III. Ofcourse other factors must also be borne in mindwhen assessing how good a particular test is: thecost, the risk of side effects, the expertise of localdiagnostic services (measures of sensitivity andspecificity are usually derived from centres ofexcellence) and the extent to which the test resultcan usefully influence management.

Rather than outline a 'stroke workup' orflowchart to be applied rigidly to all patients, wewill review each of the simple tests which are widelyavailable at present and propose what we considera cost-effective basic minimum set. For the re-mainder, we will discuss those investigations whichmay be appropriate in selected cases and then theprocedures which are research tools at present butmay have relevance in the future.

ate treatment, reduce disability and hence reducethe length of hospital stay.

Evaluation of an investigative test

The precision with which an individual test canidentify the presence or absence of disease can beexpressed in a number of measures which areassessed in a formal study comparing the testsperformance with a 'gold standard' test. For exam-ple, Doppler ultrasonography of the carotidarteries can be assessed by comparison with the'gold standard' of intra-arterial angiography todetermine sensitivity (the probability of detectingdisease when it is present) and specificity (theprobability of excluding disease when it is absent)for the detection of carotid stenosis. The positivepredictive value of a test is the proportion ofpatients with a positive test result who have thedisease and the negative predictive value is theproportion of patients with negative test resultswho do not have the disease. Sensitivity andspecificity do not change with alterations in theprevalence of disease in the population under test.However, the positive predictive value is greatly

Basic investigations applicable to every patient

These are listed in Table IV. They have remainedlargely unchanged since 1950 when the authors ofHarrison's Principles of Internal Medicine sug-gested urinalysis, determination of blood nitrogenand sugar levels, skull radiography and cerebro-spinal fluid examination.'4 The tests cost little andare in widespread use; it is perhaps not surprisingthat there have been no formal studies to assess theutility ofeach test. Many non-hospitalized patientsstill do not have these investigations performed,but perhaps should do. The aim of these tests is tosupplement the clinical search for treatable condi-tions.

Full blood count may point to rare causes ofstroke such as blood dyscrasia, polycythaemia,thrombocytopaenia, thrombocytosis or infection.

ESR-its role is to screen for vasculitis, bacterialendocarditis or hyperviscosity as the cause ofstroke.

Biochemical screen may provide evidence that theneurological deficit is due to metabolic disturbance

copyright. on N



j.bmj.com

/P


j.67.785.259 on 1 March 1991. D

ownloaded from

http://pmj.bmj.com/


Table III The ability of duplex ultrasonography of the carotids to detect moderate to severecarotid stenosis (greater than 50% decrease in diameter) in patients with differing pre-testprobabilities of disease assuming a sensitivity and specificity of approximately 95% against the

gold standard of carotid angiography

Pre-test Post-testHistory probability probability

40 year old woman with visual loss possibly 5% + ve test = > 50%migrainous, possibly vascular -ve test = > 0.3%*

45 year old man with visual loss hypertension 50% + ve test = > 95%and ?bruit - ve test =>5%

68 year old man with definite carotid transient 95% + ve test = > 99.7%tischaemic attacks, angina, peripheral vascular -ve test = > 50%disease and loud carotid bruits

*Test may be useful here to exclude presence of disease but will generate 5 false-positive resultsin every 100 patients screened (i.e. test positive but stenosis not present). tA positive test hereadds little to the clinical suspicion of disease.

Table IV Basic investigations applicable to everypatient with suspected acute stroke

Full blood count and ESRUrea and electrolytes, calciumBlood glucoseCholesterolSerological tests for syphilisChest radiographyElectrocardiograph

or electrolyte imbalance rather than a vascularlesion of the brain.

Blood glucose Hypoglycaemia and non-ketotichyperglycaemia can both be associated with strik-ing focal neurological deficits which mimic stroke,yet disappear completely when plasma glucoselevels are returned to normal. This test also helpsdiagnose occult diabetes mellitus and provides aprognostic indicator (elevated blood glucose afterstroke, even in non-diabetics, is an indicator ofadverse outcome and increased infarct size, al-though whether this reflects a cause-and-effectrelationship is unclear).'5

Serum cholesterol-if taken within 12-24 hours ofthe acute event it represents a reliable measure ofthe pre-stroke levels. This may predict occurrenceof further vascular events, and may represent a riskfactor for modification in secondary prevention. Ifthis test is not performed acutely then it should bedelayed for at least 6 weeks because of changes incholesterol level induced by the stress of the acutestroke.

Syphilis serology Meningovascular syphilis is arare cause of stroke (a frequency of 1 in 500 strokesin a recent community survey3), but one which ispotentially treatable, so screening using a non-

treponemal test such as the VDRL is probablycost-effective. 16

Chest radiography provides, with the clinical his-tory and electrocardiogram (ECG), evidence ofpotential cardioembolic lesions. In addition heartsize can indicate prolonged hypertension. Lesscommonly, unsuspected malignancy may be found,and the chest X-ray may enable early diagnosis ofcomplications such as pneumonia and pulmonaryembolism to be made.

Electrocardiography In occasional patients withstroke associated with complete heart block it isdiagnostic. Otherwise, the ECG provides confirma-tory evidence of underlying cardiovascular diseasesuch as ischaemic heart disease, arrhythmia orvalvular lesions. Prospective ECG, cardiac enzymeand autopsy studies have suggested that asymp-tomatic myocardial necrosis occurs in many pa-tients during the acute phase of stroke, althoughthe effect on outcome is unclear."7

Additional investigations in special circumstances

In general, use of these tests is guided by clinicalsuspicion or an abnormal result on one of the basictests.

(A) Haematological investigations

Serum protein levels, plasma viscosity and elec-trophoresis may be of use if hyperviscosity issuspected, usually the ESR is high in these patients.

Autoantibody screen in younger patients or inpatients with suspected autoimmune disease orvasculitis. A prolonged activated partial throm-boplastin time (APTT) or nonspecific syphilis

copyright. on N



j.bmj.com

/P


j.67.785.259 on 1 March 1991. D

ownloaded from

http://pmj.bmj.com/


serology may suggest the presence of anticar-diolipin antibodies. Even in the absence of clinicalsigns of systemic lupus erythematosus the presenceof these antibodies in high titre has been associatedwith idiopathic stroke in young patients, and mayidentify patients at particularly high risk of recur-rent vascular events.'8

Sickle test in appropriate racial groups (Ne-groes).

Haemostatic profile in patients with haemorr-hagic stroke not due to hypertension, aneurysm orarteriovenous malformation, tests of haemostasisincluding prothrombin time, activated partialthromboplastic time, thrombin time, bleeding timeor fibrin degradation products may be required. Inpatients with unexplained ischaemic stroke (es-pecially young patients or in those with a strongpast or family history of premature arterial orvenous occlusive disease), measurement of proteinS and C levels, antithrombin III levels, plateletaggregation or fibrinolytic activity may be helpful.Thrombotic disorders due to genetic abnormalitiesin the clotting system are relatively rare. Theacquired thrombotic tendency associated with'lupus anticoagulant' (anticardiolipin antibodies)is an increasingly recognized cause of unexplainedischaemic stroke,'8 and initial screening should berequested in consultation with a haematologist.Factor VIII and fibrinogen levels are associatedwith an increased risk of stroke and may predictoutcome and the risk of recurrence.'9

(B) Urine tests

Screening for homocystinuria with methionineloading, or tests for porphyria or metanephrines todetect phaeochromocytoma may be sometimesindicated.

(C) Lumbarpuncture

This investigation may occasionally be helpful byexcluding lesions mimicking stroke, such as menin-gitis, encephalitis, or to confirm subarachnoidhaemorrhage. In cases of suspected subarachnoidhaemorrhage computed tomography ofthe brain isthe preferred initial investigation.4 The ability ofcerebrospinal fluid (CSF) examination to dis-criminate between primary cerebral haemorrhageand cerebral infarction has been examined in twoprospective studies. In one, a CSF protein, mea-sured a mean of 2.5 days after stroke onset, greaterthan 1 g/l had a sensitivity of89% and a specificityof 92% for haemorrhagic stroke whilst visiblexanthochromia showed a sensitivity of 70% and aspecificity of95% where the gold standard was CTor autopsy diagnoses.20 In centres with no CT

scanning facilities, lumbar puncture (LP) and CSFexamination are still used to exclude intracerebralhaemorrhage; however, LP is clearly inferior to CTscanning particularly in the diagnosis of small deephaemorrhages and does not exclude haemorrhagesreliably enough to permit anticoagulant therapy.2'The danger of a spinal haematoma complicatingearly anticoagulation after lumbar puncture issignificant, seven cases of iatrogenic paraplegiawere reported in a series of 342 patients.2' LPshould not be performed if there is any evidence ofraised intracranial pressure.

(C) Computed tomography ofthe brain

CT scanning is very informative in stroke forseveral reasons. In the few patients for whom aclear history ofa sudden onset offocal neurologicaldeficit cannot be obtained, it differentiates 'stroke'from 'non-stroke' lesions. It can discriminatebetween haemorrhage and infarction reliably at anearly stage, and may later clarify the site and type ofinfarct and the likely underlying vascularpathology,22 with good interobserver agreement.23CT scanning would thus appear to be desirable forall cases ofstroke, but is it necessary? Sandercock etal. calculated that the cost of performing CT scanson every patient who presented with a first stroke inEngland and Wales would be at least £1.35 millionper year (1984 prices),24 without the additionalexpense of providing CT scanners to the 49% ofhealth districts and boards which do not currentlyhave such equipment.25

(1) Differentiating 'stroke' from 'non-stroke' The incidence of non-stroke lesionsdetected by CT scanning in patients with suspectedstroke ranges from 1.5-16%.2426 30 In about twothirds of patients, the detection of a 'non-stroke'lesion results in major changes to management.24'26The frequency of unexpected non-stroke lesions ina population of patients clinically suspected ofstroke can be reduced by careful elucidation of thehistory, particularly the speed of onset ofneurological deficit. The corollary is that in caseswhere no clear history is available from the patientor a reliable witness, a CT scan may be needed toconfirm the clinical diagnosis of stroke.

(2) Excluding haemorrhagic stroke CT scann-ing represents the best method of excludinghaemorrhage (CSF examination has a sensitivity ofonly 60-70% against the gold standard of CTscanning).3' During the first 24-48 hours afteronset of stroke a primary intracerebral haemorr-hage will be clearly visible as an area of markedhyperdensity provided the haemorrhage is greaterthan about 5 mm in diameter; this is probably thebest time to detect small haemorrhages. The den-

copyright. on N



j.bmj.com

/P


j.67.785.259 on 1 March 1991. D

ownloaded from

http://pmj.bmj.com/


sity of haemorrhages declines with time, so that by14-21 days the haematoma can be isodense withbrain and thus invisible, or hypodense and indistin-guishable from a small lacunar infarct.32 Infarctson early CT scan may also undergo haemorrhagictransformation at a later date in approximately20% of acute cerebral infarctions which werescanned within 6 hours of onset in one series.34 Allhaemorrhagic transformations occurred less thanone month after onset of stroke. To reliablydiscriminate between haemorrhagic and ischaemicstroke CT scanning less than 3 weeks after onset ismandatory.

In suspected cardioembolic stroke or strokewhich appears to be progressing, some cliniciansuse anticoagulant therapy, although its efficacy inaltering the natural history ofacute stroke is not yetclear. The role of antiplatelet therapy in the acutephase of ischaemic stroke from any cause and therisks of antiplatelet therapy in the presence ofhaemorrhagic infarction are uncertain, but theoverall benefit in reducing the risk of vascularevents by 25% in long term secondary preventionafter stroke is clear." A small proportion ofpatients with stroke are already on anticoagulantsor antiplatelet agents at the time of their stroke,and CT scanning is therefore needed to guidefurther treatment (stop ifCT shows haemorrhage;continue if haemorrhage excluded). An additionalneed for CT occurs when a stroke patient developsa deep venous thrombosis (DVT). Up to 50% ofparalysed legs had a DVT on fibrinogen scanningin one hospital series.36 If full-dose anticoagulationis planned in the treatment of established DVT orpulmonary embolism after stroke, CT scanning isrequired to exclude haemorrhage before treatmentis started. It is not known whether it is safe to givelow-dose subcutaneous heparin in the prophylaxisof DVT after stroke without first excluding intra-cranial haemorrhage. Finally, a small proportionof patients with minor ischaemic strokes in thecarotid distribution may be suitable for carotidendarterectomy. Since minor haemorrhagicstrokes are clinically indistinguishable from minorischaemic strokes, patients with mild strokesshould have early CT to avoid patients withhaemorrhagic strokes being inappropriately putforward for angiography and carotid surgery.

(3) Making a positive diagnosis of cerebral infarc-tion Cerebral infarction may not be apparent inthe first 24-48 hours, or the subtle area ofhypodensity induced by early cellular oedema maybe overlooked, and CT may not show an infarct inapproximately 40% of cases.29 Over the followingdays increasing oedema may result in theappearance of a mass effect with displacement ofnormal tissue. Mass effect is greater with largerinfarcts and is present to some degree in 70% of

infarcts visible on CT in the first week.37 From 1-3weeks after stroke, infiltration ofmacrophages andother inflammatory cells involved in neuronalrepair may cause the infarcted tissue to becomeisodense on CT scans, the 'fogging effect'. Thiseffect occurs in approximately 50% of infarctswhich were initially hypodense on early scans.38One way to increase the positive diagnosis ofcerebral infarction during this period is to repeatthe scan with injection ofiodinated contrast agents.This may show enhancement surrounding the areaof infarction corresponding to neovascularizationand increased permeability of the blood-brainbarrier. Contrast scanning revealed significantenhancement in all patients whose infarcts hadbecome 'fogged' at 2 weeks after ictus in oneprospective series of 59 patients,39 but providedlittle additional information in those patients whohad shown hypodense lesions on early scans. It hasalso been suggested in one retrospective study thatthe administration of contrast might adverselyaffect prognosis although the group given contrastalso had an excess of several other markers ofpoorprognosis and this work needs confirmation.' LateCT scans after several months may show hypoden-sity or localized atrophy corresponding to the areaof infarction, but may also reveal no abnormalityin up to 20% of cases despite a very clear clinicalpicture of stroke.38 The overall accuracy of CTscanning in making a positive diagnosis ofcerebralinfarction critically depends on the site and size ofthe infarct and the timing of the scan.

(4) Identifying unusual treatable forms ofstroke Some uncommon forms of stroke willbenefit from early diagnosis by CT scan, particular-ly cerebellar haemorrhage or infarction whichpresents with sudden headache, vertigo, and un-steadiness or slurred speech. In cases where cerebel-lar swelling results in obstructive hydrocephaluswith worsening neurological deficit and decliningconscious level CT scanning identifies the caseswhich may benefit from immediate neurosurgicalintervention.

Cerebral vein thrombosis presents subacutelywith headache, seizures, focal neurological deficitand altered conscious state. Early CT scan mayshow parasagittal low density areas suggestingvenous infarction, particularly if there is associatedhaemorrhage or a filling defect in the superiorsagittal sinus on axial slices with the injection ofcontrast (delta or empty triangle sign).4' Anti-coagulant therapy, even in the presence ofhaemor-rhage, has been shown to improve outcome in thelargest series of these patients reported to date.42

(5) Other indications Certain features on CTscanning such as the size of the area ofinfarction orhaemorrhage, presence of mass effect, and site of

copyright. on N



j.bmj.com

/P


j.67.785.259 on 1 March 1991. D

ownloaded from

http://pmj.bmj.com/


lesion correlate with functional outcome,4344 butprovide little additional information above thatavailable from clinical features alone. Preciseanatomical localization of infarcts is important forresearch purposes and may help if carotidendarterectomy is considered, but provides littlehelp to the average stroke patient.

A summary of the indications for CT scanning insuspected stroke is provided in Table V. Data onthe cost-effectiveness ofCT scanning and its impacton management of stroke patients are scanty.45'46 Acost-effectiveness study in Sweden found that earlyCT scanning was more cost-effective than a policyof sequential investigation including CSF analysisand radionuclide scanning, with the detection oftwo non-stroke lesions on CT in 256 patientsinvestigated which were missed with CSF andradionuclide scanning alone.47 Early CT mightreduce the cost of hospital stay by earlier diagnosisof potentially disabling yet treatable conditionsand by avoiding the occurrence of disabling cere-bral haemorrhages in patients inappropriatelytreated with anticoagulants.

(D) Diagnosis ofcardiac sources ofemboli andechocardiography

Two-dimensional echocardiography is now widelyavailable. This non-invasive investigation is rela-tively simple to perform and may identify a poten-tial cardioembolic cause of ischaemic stroke. How-ever, the application of this test to all patients withstroke would involve a huge commitment ofresources from cardiology services (in the USA, thecost would be perhaps $100 million dollarsannually').

Cardioembolic stroke is a clinical diagnosismade by excluding atherothrombotic disease in theartery supplying the area of the stroke, along withfinding a cardiac lesion likely to cause emboli to thebrain. Clinical features suggesting, but not diag-nostic of, cardiogenic embolism include:49(1) cerebral infarcts occurring in multiple vascular

territories;(2) peripheral emboli (such as femoral, mesenteric

or renal arteries) in the presence of stroke;(3) presence of cardiac thrombus on 2D echo-

cardiography;(4) presence of a potential embolic source (Table

VI);(5) absence of atherosclerotic disease in the symp-

tomatic cerebral artery (usually diagnosed withnon-invasive arterial imaging).

Pathological confirmation of cardiac embolismis rare and only in the minority of cases is a

thrombus demonstrable in the heart by presenttechniques. Current echocardiographic techniquescan detect a thrombus perhaps as small as

Table V Indications for CT scanning in patients withsuspected stroke

1) Doubt about the diagnosis of stroke because no clearhistory is available (the patient is in coma, confused ordysphasic) or the presentation is atypical (gradualonset)

2) A clinical picture of cerebellar or 'brainstem' stroke,accompanied by declining conscious level raising thepossibility of secondary obstructive hydrocephalus

3) Exclusion of intracranial haemorrhage in thosepatients already on antiplatelet or anticoagulantdrugs, or in whom the use of such drugs, thrombolyticagents or carotid endarterectomy is planned

4) Atypical progression of the stroke after onset5) Subarachnoid haemorrhage6) Young patient without obvious risk factors for stroke

Table VI Cardiac abnormalities which are associatedwith an increased risk of ischaemic stroke (echocardio-

graphy probably cost effective)

HIGH RISKRheumatic mitral stenosis and atrial fibrillationMechanical prosthetic cardiac valvesRecent large anterior transmural myocardial infarctionVentricular aneurysm with visible thrombusBacterial or marantic endocarditisCardiac myxoma

LOWER RISKMitral valve prolapse (with myxomatous degeneration)Dilated cardiomyopathy and/or cardiac failureCongenital heart disease with pulmonary hypertensionCalcific aortic or mitral valve diseaseAtrial fibrillation without valvular heart diseaseSick sinus syndrome or left bundle branch blockPatent foramen ovaleTissue heart valves

* In addition echocardiography may also be cost-effectivein patients aged less than 45 years in whom no obviouscause of stroke can be found.

5-1O mm in diameter, but a thrombus of only1-2 mm diameter is sufficient to occlude a branchof the middle cerebral artery. Furthermore,echocardiography after stroke may be 'negative' ifthe thrombus which caused the stroke has left theheart and embolized to the brain! Standard 2Dechocardiography does not visualize the atrialappendices well yet postmortem studies suggestthat this is a common site of thrombus. Thepresence of a cardioembolic source does not neces-sarily prove that the stroke is due to cerebralinfarction; in a Swedish series, 6% of patients withstroke who were in atrial fibrillation turned out, on

copyright. on N



j.bmj.com

/P


j.67.785.259 on 1 March 1991. D

ownloaded from

http://pmj.bmj.com/


CT scanning, to have a stroke due to primaryintracerebral haemorrhage.5

Problems arise in the interpretation of theclinical picture, the definition ofwhat is a potentialsource of embolism, and definition of whatrepresents atherosclerotic disease. A reasonableconsensus seems to be that somewhere between6-23% of strokes may be caused by cardiacemboli.5' It has been thought worthwhile to makethis diagnosis because a cardiac source of emboluscarries a high risk of recurrence (possibly in avascular territory remote from the initial insult),and many clinicians use urgent anticoagulation inthis situation. This is probably reasonable inpatients at very high risk of recurrence (mechanicalprosthetic valves, rheumatic mitral stenosis withatrial fibrillation, left ventricular thrombi), but inother groups at lower risk (non-valvular atrialfibrillation, tissue valves, mitral valve prolapse) therole of anticoagulation is unclear and is currentlybeing tested in trials such as the European AtrialFibrillation Trial. There is a 40% risk of haemorr-hagic transformation in cardioembolic cerebralinfarcts;3' data from current randomized trials willprovide clearer evidence on whether the presence ofhaemorrhagic transformation should be anabsolute or relative contraindication to the use ofanticoagulants.

Echocardiography represents a simple non-invasive way of identifying potential cardiacsources of emboli. The yield of echocardiographicabnormalities is much greater in patients who haveevidence ofcardiac abnormality on clinical historyand examination (supported by chest X-ray andECG). In several studies, detailed echocardio-graphy identified abnormalities associated withcardiogenic embolism in 6-30% of cases withclinical evidence of cardiac disease and only 0-3%of patients with no clinical evidence of cardiacdisease.52-56 In a retrospective study Robbins etal.56 calculated that it cost approximately $1,200US (1983 prices) for every high-risk lesionidentified in patients with known cardiovasculardisease, and that 30% of these patients had theirtreatment changed as a result of 2D echocardio-graphy. Groups in which echocardiography wouldseem cost-effective in the diagnosis of potentialcardiac sources ofembolism are listed in Table VI.

Transoesophageal echocardiography is a newtechnique currently being evaluated. An oesopha-geal probe is used to visualize the left ventricle,atria, mitral valve apparatus, aortic valve, inter-atrial septa and aortic root. It is an ideal techniquefor evaluating the atrial appendices. Recent studiessuggest that transoesophageal echocardiographyhas a much greater sensitivity than praecordialechocardiography, although the ventricular apex isusually not well seen using the former tech-

nique. 7-59 The transoesophageal approach is time-consuming and carries a small risk, since it requireslocal anaesthesia to the pharynx and sometimessedation in patients who may be restless. It maybecome necessary in some patients to perform bothpraecordial and transoesophageal techniques,greatly increasing time required and cost involved.

(E) Cerebral angiography in patients with stroke

Moniz first diagnosed occlusive cerebrovasculardisease in 1937 using the technique of carotidangiography in four patients with episodes ofweakness.' Present indications for angiography inthe acute phase of ischaemic stroke are limited tothe diagnosis of non-atherogenic cerebral vesselocclusion such as that caused by carotid arterydissection, Moya Moya disease or vasculitis, orischaemic stroke in a young patient with noobvious causative factors.

In the future, some assessment of the patency ofintra- and extra-cranial arteries may be requiredbefore treatment with thrombolytic therapy inacute ischaemic stroke. In a recent series Fieschi etal.34 demonstrated complete occlusion of theapropriate cerebral artery in 76% of patients withacute ischaemic stroke who underwent angio-graphy within 6 hours of stroke onset. Suchpatients may be suitable for entry in randomizedtrials of thrombolytic therapy, although patientswho might have most to gain would be those with aviable ischaemic 'penumbra' distal to the site ofvascular occlusion. The assessment of the site andsize of any penumbra is not feasible with currenttechniques. There were 2 cases (2.5%, 95%confidence interval 0.3-8.8%) ofclinical deteriora-tion following angiography in this series;34 angio-graphy in the acute phase therefore carries asignificant risk.Angiography may be indicated in patients who

have made a good recovery from a mild ischaemicstroke, ifcarotid endarterectomy to prevent furtherstroke is being considered. In this situation, the riskofpermanent neurological deficit as a complicationofangiography is approximately 1%.61 This comp-lication rate should be added to the 3-4% risk ofdisabling stroke/death associated with carotidsurgery. Intravenous digital subtraction angio-graphy fails to provide the anatomical definitionrequired to plan carotid surgery and carries a 17%risk of systemic complications associated with theadministration of large volumes of contrastmaterial.62 As a screening test to select patients forintra-arterial angiography, duplex ultrasoundexamination of the carotid extracranial circulationoffers a sensitivity and specificity comparable tothat of intravenous digital subtraction angio-graphy without the potential complications.63

copyright. on N



j.bmj.com

/P


j.67.785.259 on 1 March 1991. D

ownloaded from

http://pmj.bmj.com/


New investigations limited to research at present butwhich may be used in routine clinical application inthe future

(A) Transcranial Doppler ultrasonography (TCD)

This technique uses a low-frequency pulsed ultra-sound beam and a detector which can record flowvelocities at selected distances from the probe. Thebeam can be directed through an 'orbital' windowto record information about the carotid siphon, a'temporal' window to record velocities in themiddle, posterior and anterior cerebral arteries,and the foramen magnum to record velocities in thevertebrobasilar system. Absolute velocity measure-ments do not correlate well with cerebral bloodflow for a number of reasons but comparison ofvelocities with the contralateral hemisphere andassessment of collateral and retrograde flow givesinformation about proximal occlusion ofthe majorcerebral vessels. In 5-15% of patients it is notpossible to make satisfactory recordings throughthe temporal window as a result of the thickness ofthe vault or anatomical variations in the bloodvessels.' Each examination can take a long time(up to 1 hour per examination) and the diagnosticaccuracy is highly dependent on the experience ofthe operator. In future it may provide an alterna-tive to angiography in the early diagnosis andfollow-up of intracranial arterial occlusion in acuteischaemic stroke. Preliminary studies suggest ahigh sensitivity and specificity for the diagnosis ofproximal middle cerebral artery occlusion.65

(B) Nuclear magnetic resonance imaging (NMR)

This non-invasive technique records the high-frequency radio waves emitted by protons as theyare manipulated in a magnetic field. Measures ofproton density and relaxation curves allow dis-crimination between different tissue types. Tl-weighted images reflect differences in the rate oftissue magnetization and T2-weighted imagesreflect differences in the rate of signal decaybetween different tissue types. Early cerebralinfarcts (within hours of onset) appear as low signalregions on Tl-weighted images and as regions ofincreased signal with T2- weighting. NMR detectsareas of infarction, including small lacunar lesions,earlier and with greater sensitivity than CT scann-ing,6669 and may be especially helpful in theposterior fossa where conventional CT scans areoften degraded by bony artifact.70 CT scanningremains the method of choice in the detection ofcerebral haemorrhage, as NMR scans performedless than 24 hours after the onset may fail to revealthe haemorrhage.69 Enhancement of cerebralinfarcts on NMR scans occurs with the use of

gadolinium contrast during the period of neovas-cularization and blood-brain barrier breakdownin the 1-3 weeks after stroke onset.A new application of NMR scanning in acute

stroke is spectroscopy which enables measurementof ATP, lactate levels and pH at discrete locationswithin the brain. This may help identify potentiallyviable infarct tissue which could benefit fromrevascularization.7' Another NMR technique cur-rently being evaluated is angiography with digitalreconstruction of arterial flow patterns. Currenttechniques are a long way from providing picturessuperior to conventional arteriography,72'73although the potential advantages of a techniquesuch as this, which is non-invasive and requires nocontrast agent, are obvious. NMR scanning takesquite a long time to perform (30-45 minutes) andsome patients find the scanner itself claustro-phobic, both factors which reduce utility in theunwell acute stroke patient who may be alsorestless.

(C) Single photon emission computed tomography(SPECT)

This radioisotope technique measures cerebralblood flow, and, when combined with a tomo-graphic scanner, allows two-dimensional recon-struction of cerebral blood flow patterns. Thesepatterns are a functional representation of thecerebral infarct and correlate well with clinicalsigns and subsequent CT images. Tracers such as1231 iodoamphetamine or 9'technetium HMPAOgive a qualitative picture of cerebral bloodflow inthe infarct area. Changes in response tovasodilators such as acetazolamide assess the integ-rity of cerebral autoregulation and may give someidea of tissue viability.74 SPECT studies may bemore sensitive than CT scanning in the detection ofearly cerebral infarction but cannot reliably diag-nose haemorrhage or predict prognosis.75'6'33Xenon contrast CT scans are an alternativemethod of measuring cerebral blood flow butrequire the patient to remain immobile for 15-20minutes whilst the scan is performed, and xenonhas soporific effects and may alter flow by acting asa vasodilator. Xenon provides a quantitativemeasure of cerebral blood flow correlating wellwith other methods.77

(D) Positron emission tomography (PET)

Certain radio-isotopes decay spontaneously, emitt-ing a small positively charged particle (positron),which interacts with an electron, producing gammaradiation which is detectable with a gammacamera. This technique enables precise and quanti-tative analysis of radioisotope concentrations from

copyright. on N



j.bmj.com

/P


j.67.785.259 on 1 March 1991. D

ownloaded from

http://pmj.bmj.com/


molecules such as "IC, 13N, 150 and '8F which maybe attached to normal metabolic substances such asoxygen, glucose and water. Unfortunately, theseradio-isotopes can only be produced using a cyclo-tron and have short half-lives (between a fewminutes and two hours) which restricts their use toa few centres in the world. Quantitative regionalmeasurements can be made of cerebral blood flow,blood volume, oxygen uptake and glucosemetabolism enabling definition of normal brainfrom infarct, and the region of ischaemic butpotentially viable brain surrounding the infarct(the 'ischaemic penumbra').78 Areas of alteredmetabolism distant to the area of infarction havealso been identified which may have an importantpathophysiological role (diaschisis). Metabolicabnormalities on PET scanning may predict thedevelopment of future structural lesions and act asa guide to appropriate treatment in acute cerebralinfarction, although its potential for widespreaduse is limited because of the enormous costsinvolved.

Conclusion

Many investigations are now available for theassessment of patients with acute stroke. The

investigations which need to be performed in anindividual patient will depend on patient charac-teristics, management priorities and local facilities.A good clinical history and examination is the mostcost-effective investigation of all. The basic investi-gations outlined in the first section are a cost-effective addition to the clinical examination andare applicable to virtually all stroke patients. CTscanning may be required in a large proportion ofpatients, the exact proportion depending largely onwhether clinicians plan to use anti-haemostaticagents for either acute treatment or long-termsecondary prevention, whereas other investigationssuch as echocardiography are indicated only inselected patients in whom there is diagnostic uncer-tainty, or in whom management may be signi-ficantly altered. The newer research techniquesoutlined require further evaluation and mayassume great importance in the future when moreeffective therapies for the acute treatment of strokeare available.

References

1. Malmgren, R., Bamford, J., Warlow, C., Sandercock, P. &Slattery, J. Projecting the number of patients with first everstrokes and patients newly-handicapped by stroke in Englandand Wales. Br Med J 1989, 298: 656-660.

2. Isard, P.A. & Forbes, J. Stroke and health service resource usein Scotland. Personal Communication.

3. Sandercock, P.A.G., Warlow, C.P., Jones, L.N. & Starkey,I.R. Predisposing factors for cerebral infarction: the Oxford-shire community stroke project. Br Med J 1989, 298: 75-80.

4. Vermeulen, M. & VanGijn, J. The diagnosis of subarachnoidhaemorrhage. J Neurol Neurosurg Psychiatry 1990, 53:365-372.

5. Norris, J.W. & Hachinski, V.C. Misdiagnosis of stroke.Lancet 1982, i: 328-331.

6. Isaacs, B. Five years' experience of a stroke unit. HealthBulletin 1977, 35: 94-98.

7. Allen, C.M.C. Clinical diagnosis of the acute stroke syn-drome. Q J Med 1983, 52: 515- 523.

8. Barer, D.H. & Mitchell, J.R. Predicting the outcome of acutestroke: do multivariate models help? Q J Med 1989, 70:27-39.

9. Jongbloed, L. Prediction of function after stroke. Stroke1986, 17: 765-775.

10. Wade, D.T., Wood, V.A. & Hewer, R.L. Recovery afterstroke: the first three months. J Neurol Neurosurg Psychiatry1985, 48: 7-13.

11. Prescott, R.J., Garraway, W.M. & Akhtar, A.J. Predictingfunctional outcome following acute stroke using a standardclinical examination. Stroke 1982, 13: 641-647.

12. Fullerton, K.J., MacKenzie, G. & Stout, R.W. Prognosticindices in stroke. Q J Med 1988, 66: 147-162.

13. Sackett, D.L. How to read clinical journals: II. To learn abouta diagnostic test. Can Med Assoc J 1981, 124: 703-710.

14. Merritt, H.H. & Sciarra, D. Disorders of the blood vessels ofthe brain and spinal cord. In: Harrison, T.R., Beeson, P.B.,Resnik, W.H., Thorn, G.W., Wintrobe, M.M. (eds) Prin-ciples of Internal Medicine. Blakiston, Philadelphia 1950,pp. 1514-1520.

15. Kushner, M., Nencini, P., Reivich, M. et al. Relation ofhyperglycaemia early in ischaemic brain infarction to cere-bral anatomy, metabolism, and clinical outcome. Ann Neurol1990, 28: 129-135.

16. Williams, K. Screening for syphilis in pregnancy: an assess-ment of the costs and benefits. Commun Med 1985, 7: 37-42.

17. Norris, J.W. & Hachinski, V.C. Stroke units or strokecenters? Stroke 1986, 17: 360-362.

18. Levine, S.R. & Welch, K.M.A. Antiphospholipid antibodies.Ann Neurol 1989, 26: 386-389.

19. Landi, G., D'Angelo, A., Boccardi, E. et al. Hyper-coagulability in acute stroke: prognostic significance.Neurology 1987, 37: 1667- 1671.

20. Britton, M., Hultman, E., Murray, V. & Sjoholm, H. Thediagnostic accuracy of CSF analyses in stroke. Acta MedScand 1983, 214: 3-13.

21. Ruff, R.L. & Dougherty, J.H. Evaluation of acute cerebralischaemia for anticoagulant therapy: computed tomographyor lumbar puncture. Neurology 1981, 31: 736-740.

22. Bories, J., Derhy, S. & Chiras, J. CT in hemispheric ischaemicattacks. Neuroradiology 1985, 27: 468-483.

23. Shinar, D., Gross, C.R., Hier, D.B. et al. Interobserverreliability in the interpretation of computed tomographicscans of stroke patients. Arch Neurol 1987, 44: 149-155.

24. Sandercock, P.A.G., Molyneux, A. & Warlow, C. Value ofcomputed tomography in patients with stroke: OxfordshireCommunity Stroke Project. Br Med J 1985, 290: 193-197.

copyright. on N



j.bmj.com

/P


j.67.785.259 on 1 March 1991. D

ownloaded from

http://pmj.bmj.com/


25. Hewer, R. & Wood, V.A. Availability of computed tomo-graphy of the brain in the United Kingdom. Br Med J 1989,298: 1219-1220.

26. Matenga, J., Kitai, I. & Levy, L. Strokes among black peoplein Harare, Zimbabwe: results of computed tomography andassociated risk factors. Br Med J 1986, 292: 1649-1651.

27. Britton, M., Hindmarsh, T., Murray, V. & Tyden, S.A.Diagnostic errors discovered by CT in patients withsuspected stroke. Neurology 1984, 34: 1504-1507.

28. Weisberg, L.A. & Nice, C.N. Intracranial tumours simulatingthe presence of cerebrovascular syndromes: early detectionwith cerebral computed tomography. Am J Med 1977, 63:517-524.

29. Campbell, J.K., Houser, O.W., Stevens, J.C., Wahner, H.W.,Baker, H.L. & Folgar, W.N. Computed tomography andradionuclide imaging in the evaluation of ischaemic stroke.Radiology 1978, 126: 695-702.

30. Kinkel, W.R. & Jacobs, L. Computerized axial transversetomography in cerbrovascular disease. Neurology 1976, 26:924-930.

31. Weisberg, L.A. Computerized tomography in intracranialhaemorrhage. Arch Neurol 1979, 36: 422-426.

32. Dennis, M.S., Bamford, J.M., Molyneux, A.J. & Warlow,C.P. Rapid resolution of signs of primary intracerebralhaemorrhage in computed tomograms of the brain. Br MedJ1987, 295: 379-381.

33. Okada, Y., Yamaguchi, T., Minematsu, K. et al. Haemor-hagic transformation in cerebral embolism. Stroke 1989, 20:598-603.

34. Fieschi, C., Argentino, C., Lenzi, G.L., Sacchetti, M.L., Toni,D. & Bozzao, L. Clinical and instrumental evaluation ofpatients with ischaemic stroke within the first six hours. JNeurol Sci 1989, 91: 311-322.

35. Antiplatelet Trialists' Collaboration. Secondary preventionof stroke, myocardial infarction and vascular death byprolonged antiplatelet treatment. J Neurol 1990, 237: 130.

36. Warlow, C., Ogston, D. & Douglas, A.S. Deep venousthrombosis of the legs after strokes. Part 1 - Incidence andpredisposing factors. Br Med J 1976, 1: 1178-1183.

37. Masdeu, J.C., Azar-Kia, B. & Rubino, F.A. Evaluation ofrecent cerebral infarction by computed tomography. ArchNeurol 1977, 34: 417-421.

.38. Skriver, E.B. & Olsen, T.S. Transient disappearance ofcerebral infarcts on CT scan, the so-called fogging effect.Neuroradiology 1981, 22: 61-65.

39. Skriver, E.B. & Olsen, T.S. Contrast enhancement of cere-bral infarcts: incidence and clinical value in different states ofcerebral infarction. Neuroradiology 1982, 23: 259-265.

40. Kendall, B.E. & Pullicino, P. Intravascular contrast injectionin ischaemic lesions. II. Effect on prognosis. Neuroradiology1980, 19: 241-243.

41. Banerjee, A.K. & Carvalho, P. An unusual cause for a fit.Postgrad Med J 1990, 66: 842-843.

42. Einhaupl, K.M., Villringer, A., Haberl, R.L., Dirnagl, U.D.,Pfister, H.W. & Schmiedek, P. Diagnosis of sinus venousthrombosis. J Neurol 1990, 237: 159.

43. Valdimarsson, E., Bergvall, U. & Samuelsson, K. Prognosticsignificance of cerebral computed tomography results insupratentorial infarction. Acta Neurol Scand 1982, 65:133-145.

44. Brott, T. & Reed, R.L. Intensive care for acute stroke in thecommunity hospital setting: the first 24 hours. Stroke 1989,20: 694-697.

45. Hazelton, A.E. & Earnest, M.P. Impact of computed tomo-graphy on stroke management and outcome. Arch InternMed 1987, 147: 217-220.

46. Larson, E.B., Omenn, G.S. & Loop, J.W. Computed tomo-graphy in patients with cerebrovascular disease: impact of anew technology on patient care. Am J Roentgenol 1978, 131:35-40.

47. Britton, M., Jonsson, E., Marke, L.A. & Murray, V. Diag-nosing suspected stroke: a cost-effectiveness analysis. Int JTech Assess 1985. 1: 147-158.

48. Larson, E.B., Stratton, J.R. & Pearlman, A.S. Selective use oftwo-dimensional echocardiography in stroke syndromes.Ann Intern Med 1981, 94: 112-113.

49. Cerebral Embolism Task Force. Cardiogenic brain embo-lism. Arch Neurol 1986, 43: 72-81.

50. Britton, M. & Gustafsson, C. Non-rheumatic atrial fibrilla-tion as a risk factor for stroke. Stroke 1985, 16: 182- 187.

51. Cerebral Embolism Task Force. Cardiogenic brainembolism: the second report of the cerebral embolism taskforce. Arch Neurol 1989, 46: 727-743.

52. Greenland, P., Knopman, D.S., Mikell, F.L., Asinger, R.W.,Anderson, D.C. & Good, D.C. Echocardiography in diag-nostic assessment of stroke. Ann Intern Med 1981, 95: 51 - 53.

53. Knopman, D.S., Anderson, D.C., Asinger, R.W., Greenland,P., Mikell, F. & Good, D.C. Indications for echocardio-graphy in patients with ischaemic stroke. Neurology 1982, 32:1005-1011.

54. Burnett, P.J., Greenwood, R., Milne, J.R., Giles, M.R. &Camm, J. The role of echocardiography in the investigationof focal cerebral ischaemia. Postgrad Med J 1984, 60:116-119.

55. Lovett, J.L., Sandok, B.A., Giuliani, E.R. & Nasser, F.N.Two-dimensional echocardiography in patients with focalcerebral ischaemia. Ann Intern Med 1981, 95: 1-4.

56. Robbins, J.A., Sagar, K.B., French, M. & Smith, P.J.Influence of echocardiography on management of patientswith systemic emboli. Stroke 1983, 14: 546-549.

57. Voll, C.L., Polasek, P., Cujec, B. & Shuaib, A. Tran-soesophageal echocardiography improves diagnosis ofpotential cardioembolic sources in unselected stroke patients.Can J Neurol Sci 1990,17:229-Ab 66.

58. Pop, G., Sutherland, G.R., Koudstaal, P.J., Sit, T.W., DeJong, G. & Roelandt, R.T.C. Transoesophageal echocardio-graphy in the detection of intracardiac embolic sources inpatients with TIAs. Stroke 1990, 21: 560-565.

59. Stollberger, C., Brainin, M., Slany, J. & Wien, R. Detection ofparadoxical embolism by combined use of transoesophagealechocardiography (TEE) and phlebography. J Neurol 1990,237: 144.

60. Gurdjian, E.S. History of occlusive cerebrovascular disease:II After Moniz, with special reference to surgical treatment.Arch Neurol 1979, 36: 427-432.

61. Hankey, G.J., Warlow, C.P. & Sellar, R.J. Cerebral angio-graphic risk in mild cerebrovascular disease. Stroke 1990, 21:209-222.

62. Cebul, R.D. & Paulus, R.A. The failure ofintravenous digitalsubtraction angiography in replacing carotid arteriography.Ann Intern Med 1986, 104: 572-574.

63. Tans, J.T.J. & Hoogland, P.H. The role of digital subtractionangiography of the carotid bifurcation in the evaluation ofpatients with reversible ischaemic attacks or stroke. Stroke1985, 16: 435-440.

64. Caplan, L.R., Brass, L.M., DeWitt, L.D. et al. TranscranialDoppler ultrasound: present status. Neurology 1990, 40:696-700.

65. Zanette, E.M., Fieschi, C., Bozzao, L. et al. Comparison ofcerebral angiography and transcranial Doppler sonographyin acute stroke. Stroke 1989, 20: 899-903.

66. Hommel, M., Besson, G., Le Bas, J.F. et al. Prospective studyof lacunar infarction using magnetic resonance imaging.Stroke 1990, 21: 546-554.

67. Brant-Zawadzki, M., Weinstein, P., Bartkowski, H. &Moseley, M. MR imaging and spectroscopy in clinical andexperimental cerebral ischaemia: a review. Am J Roentgenol1987, 148: 579-588.

68. Ramadan, N.M., Deveshwar, R. & Levine, S.R. Magneticresonance and clinical cerebrovascular disease: an update.Stroke 1989, 20: 1279-1283.

69. Kertesz, A., Black, S.E., Nicholson, L. & Carr, T. Thesensitivity and specificity of MRI in stroke. Neurology 1987,37: 1580-1585.

copyright. on N



j.bmj.com

/P


j.67.785.259 on 1 March 1991. D

ownloaded from

http://pmj.bmj.com/


70. Davis, S.M., Tress, B.M., Dowling, R., Donnan, G.A., Kiers,L. & Rossiter, S.C. Magnetic resonance imaging in posteriorcirculation infarction: impact on diagnosis and management.Aust NZ J Med 1989, 19: 219-225.

71. Felber, S.R., Aichner, F.T. & Sauter, R. Sequential IHMR-spectroscopy: new insights into ischaemic mechanisms.J Neurol 1990, 237: 156-157.

72. Haacke, E.M. & Masaryk, T.J. The salient features of MRangiography. Radiology 1989, 173: 611-612.

73. Ross, J.S., Masaryk, T.J., Modic, M.T., Harik, S.I., Wiz-nitzer, M. & Selman, W.R. Magnetic resonance angiographyof the extracranial carotid arteries and intracranial vessels: areview. Neurology 1989, 39: 1369-1376.

74. Raynaud, C., Rancurel, G., Tzourio, N. et al. SPECTanalysis of recent cerebral infarction. Stroke 1989, 20:192-204.

75. Hayman, L.A., Taber, K.H., Jhingran, S.G., Killian, J.M. &Carroll, R.G. Cerebral infarction: diagnosis and assessmentof prognosis by using 123IMP-SPECT and CT. AJNR 1989,10: 557-562.

76. Fazekas, F., Valetitsch, H., Schmidt, R. et al. Imagingconsiderations for the early detection of cerebral ischaemiclesions. J Neurol 1990, 237: 147.

77. Hughes, R.L., Yonas, H., Gur, D. & Latchaw, R. Cerebralblood flow determination within the first 8 hours of cerebralinfarction using stable xenon-enhanced computed tomo-graphy. Stroke 1989, 20: 754-760.

78. Powers, W.J. & Raichle, M.E. Positron emission tomographyand its application to the study of cerebrovascular disease inman. Stroke 1985, 16: 361-376.

copyright. on N



j.bmj.com

/P


j.67.785.259 on 1 March 1991. D

ownloaded from

http://pmj.bmj.com/

Documents

Decisions Investigation stroke: strategy?