1521

of their graduates, who are required to show diverse skills and abilities. J. A. Ryles 1931 description is valid today, and will continue to be so in 2021; and, in posteritys harsh judgment, doctors will continue to be found wanting on at least the last of Ryles attributes: "There is probably no servant of the community of whom a greater degree of omniscience is demanded, or upon whom a graver responsibility in respect of personal and sometimes social imposed. As a rule, differing but little in guidance is endowments and early training from others who are native given the advantages of a higher education, he is nevertheless expected to combine in his person the attributes of scientist, healer, priest and prophet".2

expects much of medical schools, and

even more

...

REFERENCES
1. McGuire C. The curriculum for the year 2000. Med Educ 1989; 23: 221-27. 2. Parry KM. The curriculum for the year 2000. Med Educ 1989; 23: 301-04. 3. Churchland P. Cognitive activity in artificial neuronal networks. In: Osherson DN, Smith EE, eds. Thinking: an invitation to cognitive science, vol 3. Cambridge, Mass: MIT Press, 1990: 199-227. 4. Hsu F-H, Anantharaman T, Campbell M, Nowatzyk A. A grandmaster chess machine. Sci Am 1990; 263, no 4: 18-24. 5. Schwartz S, Griffin T. Medical thinking: the psychology of medical judgment and decision making. New York: Springer-Verlag, 1986:

education. Med Educ 1988; 22: 456-67. 7. Gilbert W. Towards a paradigm shift in biology. Nature 1991; 349: 99. 8. Lomas J, Anderson GM, Domnick-Pierre K, et al. Do practice guidelines guide practice? The effect of a consensus statement on the practice of physicians. N Engl J Med 1990; 321: 1306-11. 9. Editorial. Research in conflict with teaching. Lancet 1987; ii: 1128. 10. Coles CR. Evaluating the effects curricula have on student learning: toward a more competent theory for medical education. In: Norman ZM, Schmidt HG, Ezzat ES, eds. Innovation in medical education: an evaluation of its present status. New York: Springer, 1990: 76-87. 11. Walton HJ, Matthews MB. Essentials of problem-based learning. Med Educ 1989; 23: 542-58. 12. Flexner A. Medical education in the United States and Canada: a report for the Carnegie Foundation for the Advancement of Teaching, 1910. 13. McManus IC, Wakeford RE. A core medical curriculum. Br Med J 1989; 298: 1051. 14. Nielsen J. Hypertext and hypermedia. New York: Academic Press, 1990. 15. Helsel SK, Roth JP, eds. Virtual reality: theory practice and promise, London: Meckler, 1990. 16. Stix G. Reach out: touch is added to virtual reality simulations. Sci Am 1991; 264: no 2 (February), p 116. 17. Van der Vleuten CPM, van Luyk SJ. Evaluating undergraduate training in medical skills. In: Nooman ZM, Schmidt HG, Ezzat ES, eds.

18. 19.

20.

198-213.

21. B.

6. Clayden GS, Wilson

Computer-assisted learning

in medical

Innovation in medical education: an evaluation of its present status. New York: Springer, 1990: 404-21. Bloom SW. The medical school as a social organisation: the sources of resistance to change. Med Educ 1989; 23: 228-41. Grant J, Gale R. Changing medical education. Med Educ 1989; 23: 252-57. Kantrowitz M, Kaufman A, Mennin S, Fülöp T, Guilbert J-J. Innovative tracks at established institutions for the education of health personnel. Geneva: WHO, 1987. Ryle JA. The natural history of disease, 2nd ed. London: Oxford University Press, 1948: 1-2.

CLINICAL PRACTICE
Classification and natural history of clinically identifiable subtypes of cerebral infarction

We describe the incidence and natural history of four clinically identifiable subgroups of cerebral infarction in a community-based study of 675 patients with first-ever stroke. Of 543 patients with a cerebral infarct, 92 (17%) had large anterior circulation infarcts with both cortical and subcortical involvement (total anterior circulation infarcts, TACI); 185 (34%) had more restricted and

groups. Those in the POCI group were at greater risk of a recurrent stroke later in the first year after the

predominantly cortical infarcts (partial anterior circulation infarcts, PACI); 129 (24%) had infarcts clearly associated with the vertebrobasilar arterial territory (posterior circulation infarcts, POCI); and 137 (25%) had infarcts confined to the territory of the deep perforating arteries (lacunar infarcts, LACI). There were striking differences in natural history
between the groups. The TACI group had a negligible chance of good functional outcome and

index event but had the best chance of a good functional outcome. Despite the small anatomical size of the infarcts in the LACI group, many patients remained substantially handicapped. The findings have important implications for the planning of stroke treatment trials and suggest that various therapies could be directed specifically at the

subgroups.
Introduction

Any classification of cerebral infarction (CI) should be based on a detailed understanding of the pathophysiological mechanisms operating in each patient so that new
ADDRESSES Department of Neurology, St Jamess University Hospital, Leeds (J. Bamford, MRCP); Department of Clinical Neurosciences, Western General Hospital, Edinburgh (P. Sandercock, MRCP, M Dennis, MRCP, C. Warlow, FRCP); and Rivermead Rehabilitation Centre, Oxford, UK (J. Burn, MRCP) Correspondence to Dr J Bamford, Department of Neurology, St Jamess University Hospital, Beckett Street, Leeds LS9 7TF, UK.

mortality was high. More than twice as many deaths were due to the complications of immobility than to direct neurological sequelae of the infarct. Patients in the PACI group were much more likely to have an early recurrent stroke than were patients in other

1522

then be targeted to and tested on specific types of infarction. However, if large epidemiological studies, which are necessary to plan clinical trials, or the trials themselves are to be feasible, a simple and widely 1 A applicable method of classification is needed. classification based on pathophysiology requires extensive investigations, which are time consuming and may not be available to many practitioners. For trials in acute ischaemic stroke in which there may be a narrow therapeutic "time window" and in community-based epidemiological studies in which many patients may not be admitted to hospital, such detailed investigation is impracticable. Furthermore, even after detailed investigation, the pathophysiological mechanism in as many as 40% of infarcts remains undefmed.2 Epidemiologists and workers involved in clinical trials have approached this dilemma in various ways. Some investigators have chosen not to have any major subdivisions, although patients with a poor prognosis might be excluded.3 Others have examined the importance of specific treatments in subgroups defined by the presence of a certain factor (eg, internal carotid occlusion4), even though the exact pathophysiological mechanism that caused the index neurological event was unknown. A few investigators have attempted to subdivide cases according to the site and size of the infarct (eg, the Canadian American Ticlopidine Study in Thromboembolic Stroke); this method, especially if it is based on clinical features, is rapid and can be applied to most patients. However, there are no data on the natural history of such subgroups from community-based epidemiological studies. Our hypothesis is that, with simple clinical criteria, a few subgroups of CI with certain distinctive characteristics, which are of value in clinical research and practice, can be defined. Our aim was not to develop a classification that disregarded the results of more detailed investigations, but rather to describe a framework within which individual physicians could use those investigations available to them to refine our primary clinically-based divisions. Such a scheme has the advantage that it is quick and practical, and it can be used by all physicians dealing with acute stroke. A second hypothesis, which we did not directly test, is that owing to the composition of the subgroups the underlying pathological processes are moderately homogeneous and therefore specific treatments might be directed towards particular subgroups. We here present data about the natural history of patients with four easily defined subgroups of CI from the Oxfordshire Community Stroke
treatments can

led

to

death, with

no

apparent

cause

other than that of vascular

origin.8
Definition of cerebral infarction

(CI)

Patients had definite CI if (a) a CT scan done within 28 days of the onset of symptoms showed an area of low attenuation, no relevant abnormality, or an area of irregular high attenuation within a larger area of low attenuation (ie, an area of haemorrhagic infarction), or if (b) a necropsy examination showed an area of CI (pale or haemorrhagic) in a region compatible with the clinical symptoms and signs. Patients with clinically definite first stroke who had not undergone computed tomography within 28 days of onset of symptoms or those in whom an adequate necropsy examination had not been done but in whom the Guys Hospital Stroke Diagnostic Scale (GHSDS) score was less than 4, were regarded as cases of probable CI. The GHSDS uses a combination of clinical signs and symptoms to give a probability that a stroke is due to CI or intracranial haemorrhage.9 This system is more accurate than an unstructured clinical diagnosis but we have included only patients in whom the probability of CI was greater than 90%. In the following analyses cases of definite and probable CI are combined.

Definitions of CI subtypes
Patients with CI were allocated to one of four groups according to It is important to emphasise that we used the clinical pattern at the time of maximum deficit from first cerebrovascular event to classify the patient and not necessarily the pattern at the time of our examination when some features might have resolved (although in most cases the two were the same). Lacunar infarcts (LA C/)-These patients presented with a pure motor stroke, pure sensory stroke, sensori-motor stroke, or ataxic hemiparesis." Although patients with faciobrachial and brachiocrural involvement were included, those with more restricted deficits were not. There were no cases of acute focal movement disorders due to CI in our study although there is some evidence that such cases should be considered in this group." Total anterior circulation infarcts (TA C/)-Patients with TACI presented with the combination of new higher cerebral dysfunction

presenting symptoms and signs.

(eg, dysphasia, dyscalculia, visuospatial disorder); homonymous visual field defect; and ipsilateral motor and/or sensory deficit of at least two areas of the face, arm, and leg. If the conscious level was impaired and formal testing of higher cerebral function or the visual fields was not possible, a deficit was assumed. Partial anterior circulation infarcts (PA C/)-Patients presented with only two of the three components of the TACI syndrome, with higher cerebral dysfunction alone, or with a motor/sensory deficit more restricted than those classified as LACI (eg, confined to one limb, or to face and hand but not to the whole arm). Posterior circulation infarcts (POC/)- These patients presented with any of the following: ipsilateral cranial nerve palsy
with contralateral
motor and/or sensory deficit; bilateral motor deficit; disorder of conjugate eye movement; cerebellar dysfunction without ipsilateral long-tract deficit (ie, ataxic hemiparesis); or isolated homonymous visual field

Project (OCSP)--a large, prospective, community-based study of first-ever stroke.

Patients and methods
The OCSP methods and characteristics of the population have been described previously.6,7 Briefly, all patients in a population of about 105 000 who presented during a four-year period with their first-ever stroke were registered with the study. The patients were assessed as soon as possible after the event by a study neurologist irrespective of whether they were admitted to hospital. A standard neurological history and examination was recorded although special attention was paid to any symptoms that had been present initially but had resolved by the time of examination. We attempted to obtain a computed tomographic (CT) scan or a necropsy examination in all cases. A diagnosis of stroke was made if there were rapidly developing clinical symptoms or signs of focal, and at times global (for patients in deep coma and those with subarachnoid haemorrhage), loss of cerebral function, with symptoms that lasted more than 24 h or that

and/or
defect.

sensory

Anatomical basis of the subdivision and

pathophysiological hypotheses
LACI-The four major lacunar syndromes are highly predictive of a small lacunar infarct in the basal ganglia or the pons.lO In most cases, when computed tomography is negative, magnetic resonance imaging* shows a lacune in the pons. There are occasional reports of non-lacunar infarcts presenting with lacunar syndromes but whether it was the maximum deficit from a single stroke that was reported is uncertain; that such lesions seemed mostly to involve the right hemisphere,1o,12 may reflect the relative insensitivity of standard clinical examination to detect higher cortical dysfunction in this area. Most lacunar infarcts are thought to be caused by intrinsic disease of a single basal perforating artery, either by lipohyalinosis or by microatheroma.13

1523

LACI = lacunar infarcts; TACI =total anterior circulation mfarcts PACI = partial anterior circulation Infarcts, POCI posterior circulation Infarcts.
=

TAC/-This

term

is

not meant to

be synonymous with

transhemispheric infarction; rather this is the clinical syndrome in which there has been ischaemia in both the deep and superficial territories of the middle cerebral artery (MCA). TACIs do not distinguish those in which the anterior cerebral artery (ACA) territory is also infarcted. Allen et al14 have shown that the volume of the infarct in patients with the complete clinical syndrome is significantly greater than that in patients with more restricted deficits (ie, LACI or PACI in our classification). There is substantial necropsy and radiological support that such infarcts only occur if the proximal stem of the MCA is occluded either by embolism or by spread of thrombus into this vessel from a more proximal occlusion, usually of the internal carotid artery (ICA).1S
Escourelle and Poirier6 estimated that the ratio of these two mechanisms (embolism/spread of thrombus) is about 2 to 1. The heart has been suggested as the most common source of embolic material but satisfactory data are lacking. PAC/-This group consists of the more restricted cortical infarcts due to occlusion of the upper division of the MCA (when there is usually no visual field deficit) or the lower division (when the motor/sensory deficit is often negligible), in addition to individual branch occlusions." PACI also includes isolated infarcts in the ACA territory and striatocapsular infarctions which usually result from a thrombus in the MCA stem overlying the origins of some of the lenticulostriate branches while the leptomeningeal collaterals maintain the viability of the overlying cortex.17 Most of the distal branch occlusions are likely to be due to embolism-either artery-to-artery or cardiogenic. However, infarction in the

areas of infarction with the occlusion of specific vessels.18 Therefore, we have considerd infarcts that, clinically, were

correlate the

associated with the brainstem, cerebellum, or occipital lobes together. There is substantial pathological evidence to suggest that the ratio of in-situ thrombosis to embolism is about 4 to 1.16 Embolism is thought more likely to lead to pure occipital lobe syndromes and the "top of the basilar" syndrome.19 Non-embolic infarction may be due to atheroma in the parent artery overlying the origin of a circumferential atery, to atheromatous occlusion of individual vessels, or to failure of collateral supply after a more proximal occlusion.

Patient grouping

frontoparietal boundary zone area might be due to more proximal occlusion where perfusion of the deep perforating vessels is maintained. The relative frequency of the different mechanisms is
unknown.

POC/-Owing to the variable vascular anatomy and extensive collateral channels around the brainstem, it is often difficult to

Although most of our patients underwent computed tomography necropsy, we would like to emphasise that for this report these investigations have been used solely to distinguish CI from primary intracerebral haemorrhage and not to move patients from one clinically defmed subgroup of CI to another. The use of computed tomography to refine the clinically identified groups will be reported elsewhere. The only exceptions were the patients who had incompletely recorded symptoms or signs at presentation but in whom the diagnosis of cerebral infarction was made by computed tomography or necropsy. Most of these patients died before they could be assessed by a study neurologist, often because they had had a stroke when they were away from the study area. In most cases it is likely that if a study neurologist had seen the patient immediately after the stroke (as would happen in usual clinical practice), he would have been able to allocate that patient to one of the clinically defmed groups. Exclusion of such patients might have led to bias in favour of a better prognosis and, therefore, in these few patients only, we have used the computed tomography or necropsy findings as the primary way of allocating them to one of the subgroups depending on the site and size of the area of infarction. Patients who were recruited during the first three years of the study were grouped according to details from the standardised
or

TABLE il-CASE FATALITY RATES AND FUNCTIONAL STATUS

Fig 1-Age-specific incidence rates.

LACI (0), TACI PACI (8), POCI

given as no of patients (%). Dep=funcUonally dependent (Rankm grades 3-5) Ind = functionally mdependent (Rankin grades 0-2)
Data
are

(.).

Other abbreviations

as

for table

i.

1524

TABLE III-CAUSES OF DEATH WITHIN 30 DAYS OF FIRST

TABLE IV-RISK OF RECURRENT STROKE

STROKE

Data

are

hazard ratios

(95% confidence intervals)

*p<005
Data are no of deaths (%) *Excludes 1 death after a recurrent stroke Abbreviations as for tableI

records of the OCSP. This grouping was done without knowledge of the details of the CT scan, necropsy, or data about functional outcome. During the fourth year, patients were grouped at the time of first examination by a study neurologist. All patients were studied prospectively and none was lost to follow-up. Our research nurses interviewed survivors 1, 6, and 12 months after the stroke. Some simple rating scales were used to measure disability and handicap. To describe functional outcome we will use two composite groups derived from the modified Rankin Scale;20 the two groups were (a) capable of independent existence or (b) dependent on others to some extent. The questionnaire was also designed to detect recurrent cerebrovascular and cardiovascular events. If a recurrent stroke was suspected, the study neurologist re-examined the patient. If a patient died, all hospital and general practitioner records relating to the death were examined. The cause of death within 30 days of onset was defmed as previously described.21 Briefly, based on all available evidence, deaths were attributed to direct neurological sequelae of the first or recurrent strokes, to the complications of immobility such as pneumonia and pulmonary embolism, to cardiac disease, or to other causes.

The presenting clinical syndromes were LACI (133), TACI (85), PACI (183), POCI (124), and unknown (20). We assigned with confidence all but 2 of the 20 patients with unknown clinical syndromes to one of the other four groups,
10 based on a necropsy and 8 on computed tomography. The reasons for uncertainty about the presenting clinical syndrome in these cases were co-existing dementia (4), recurrent stroke (4), or death (6) before assessment by the study neurologist, and imprecise clinical details in 4 patients who had a stroke while they were away from the study area. Therefore, a total of 543 patients were included in the analyses (ie, 2 less than in our previous report) (table l). Age-specific incidence rates are shown in

fig 1.
The overall case fatility rate was 10% (56/543) at 30 days, (99/543) at 6 months, and 23% (124/543) at 1 year post-stroke. Furthermore, at these times, of the incident cases, 39% (214/543),29% (159/543), and 28% (150/543), respectively, were functionally dependent. Outcome by subgroup is shown in table II and the causes of the deaths within 30 days of onset are shown in table ill. All 15 deaths due to the direct neurological sequelae of the first stroke occurred within 7 days of onset. If day 1 is taken as the day of the stroke, the deaths in the TACI group occurred on days 1 (1case), 2 (2), 3 (2), 5 (1), 6 (2), and 7 (1), whereas those in the POCI group occurred on days 2 (3), 3 (1), and 4 (2). There were no neurological deaths in the LACI or PACI groups. 86% (25/29) of deaths due to complications of immobility occurred after 7 days. In the first year, 14% (76/543) of patients had a fatal or non-fatal recurrent stroke. The rates among the subgroups were LACI 9% (12/137), TACI 6% (6/92), PACI 17% (32/185), and POCI 20% (26/129). There were three different patterns of recurrence as evidenced by the timing (fig 2) and the hazard ratios at various times post-stroke

18%

Statistics
Hazard ratios and 95 % confidence intervals were calculated with standard formulae.

Results
675 patients were registered with the OCSP, of whom, 545 (81 %) had cerebral infarcts. This diagnosis was confirmed in 439 patients (65%) by computed tomography within 28 days of onset or necropsy and in 106 (16%) by GHSDS score. 70 of these 106 patients also had had a CT scan, though more than 28 days from onset.

(table iv)&mdash;a pronounced early recurrence rate (PACI), a moderately high early recurrence rate with further episodes throughout the first year (POCI), and low even recurrence rate throughout the year (LACI).
Discussion
The value of defining subgroups of stroke patients primarily by clinical criteria is seen, for example, in subarachnoid haemorrhage (SAH). The diagnosis of SAH has a hierarchical structure; the simplest and most sensitive method of identifying patients-namely, the clinical history and examination-comes before the other more specific and invasive tests, such as computed tomography, lumbar puncture, and angiography. Thus, even if the pathophysiological mechanism responsible for the SAH is not identified (for example, aneurysmal rupture versus arteriovenous malformation), large trials of treatments,12 as well as community-based epidemiological studies,? can be done. By contrast, in first-ever stroke due to CI, the usual

Fig 2-Time to first

Abbreviations
as

recurrent

stroke.

for tablei

1525

diagnostic classifications,

such as the International Classification of Diseases of the World Health Organisation

standard clinico-anatomical correlations

are

likely

to

be

adequate in most cases.

Our results show that the four groups do have distinctive features. In the TACI group the chance of a good functional outcome is negligible and mortality is high. As expected, all the deaths due to direct neurological sequelae of stroke in the anterior circulation (ie, transtentorial herniation) occurred in this group. By contrast, although patients in the PACI group did not die from neurological sequelae, they were much more likely to have an early recurrent stroke than were patients in other groups, which will tend to lead to an accruing neurological deficit. This pattern would be compatible with the notion of an active embolic source-for example, the extracranial internal carotid artery-releasing showers of emboli for a short period but then becoming quiescent on endothelialisation of the lesion. The observations of the LACI group show that the commonly held view that such strokes are "mild" is erroneous. Although they have a low case fatality rate, a large proportion are substantially handicapped, which therefore justifies attempts to direct specific treatments at this group. The pattern of recurrent strokes is compatible with the hypothesis that for a further stroke to develop another perforating artery has to become symptomatic-ie, the strokes are unlikely to be due to an active embolic source. Finally, in the POCI group there are a few patients who die early presumably due to interference with vital brainstem structures. That these patients are probably admitted to hospital preferentially over the milder cases may have led to previous reports that prognosis in vertebrobasilar infarction was poor. This group also has a significant risk of
recurrence.

(WHO), use only a pathophysiological approach. Necropsy studies have shown that many pathophysiological mechanisms can lead to CI (eg, in-situ arterial thrombosis, cardiogenic and artery-to-artery embolism, intrinsic small vessel vasculopathies, arteritides, and global hypoperfusion) and, there is no convincing evidence that such processes can usually be distinguished by their clinical features. Although a "potential" mechanism may be demonstrated in a particular patient, a defmite pathogenetic association is difficult to establish,23 and sometimes potential sources of embolism are not detected in life (eg, cardiogenic embolism24). The recent WHO Task Force report25 recognised that a
subdivision based on site and size of the infarct should also be considered. Since computed tomography is needed to distinguish reliably between CI and primary intracerebral

haemorrhage,26 a subgrouping of CI based on this procedure has many attractions. However, there are several drawbacks to this method. First, if computed tomography is done in the acute phase, up to 50% of scans fail to show any abnormality.27 Second, there is no evidence that measurement of the volume of low-density lesions is of better prognostic value than the clinical examination, and the lesions change substantially in volume over time.28 Third, the dating of infarcts by computed tomography is unreliable and there is ample evidence that lesions that are not related to the presenting symptoms may often be present.29 Thus, patients might be grouped inappropriately. Many of the above drawbacks apply to MRI and, for various reasons, a significant minority of patients cannot be scanned.l1 Another difficulty with the classification of patients according to radiological appearances is the variation in the spatial resolution of scanning equipment. Therefore, the results of studies in which an imaging technique is used as the primary way of defming subgroups are not always comparable because of the difference in
scanners.

findings have important implications for the planning of stroke treatment and secondary prevention trials, and suggest that various therapies could be better directed at subgroups of CI, thereby reducing the chance of missing a clinically useful benefit because of lack of response
Our
in other groups. The clinical classification needs to be investigated further to try to establish whether the hypotheses about the underlying vascular lesions are
correct.

There is nothing new about classifying strokes according the presenting clinical syndromes. Our aim was to find a balance between the detailed classifications (in which any single type would occur so infrequently that they would be of little practical use) and the very simple classifications, such as the anterior/posterior circulation division, which may fail to recognise clinically useful subgroups. We hoped also that the classification would identify groups of patients in whom, on the basis of previous pathological studies, one of the several possible pathophysiological mechanisms might predominate. Although our study does not test this hypothesis directly, patients with strokes due to different pathophysiological mechanisms might have a different prognosis. The only previous attempt to subdivide CI along similar lines in a community-based series of first-ever stroke distinguished hemisphere and vertebrobasilar distribution strokes.3O Our classification differs in two important ways. First, it separates off the group of lacunar infarcts. How LACIs are grouped in other classifications is difficult to assess because lesions in the basis pontis and the basal ganglia can have identical motor and sensory presentations, and patients with ataxic hemiparesis syndrome (from both supratentorial and infratentorial lesions) would probably always be counted as vertebrobasilar distribution infarcts. Second, infarcts involving the cortical distribution of the carotid artery are divided. Although we recognise that there are going to be exceptions to any clinical "rules", the
to

REFERENCES
1. Sterman AB, Furlan AJ, Pessin M, Kase C, Caplan L, Williams G. Acute stroke therapy trials: an introduction to recurring design issues. Stroke 1987; 18: 524-27. 2. Sacco RL, Ellenberg JH, Mohr JP, et al. Infarcts of undetermined cause: the NINCDS Stroke Data Bank. Ann Neurol 1989; 25: 382-90. 3. Barer DH, Cruickshank JM, Ebrahim SB, Mitchell JRA. Low dose beta-blockade in acute stroke ("BEST trial"): an evaluation. Br Med J 1988; 296: 737-41. 4. Extracranial-Intracranial Bypass Study Group. Failure of extracranialintracranial bypass to reduce the risk of ischemic stroke: results of an international randomized study. N Engl J Med 1985; 313: 1191-200. 5. Gent M, Blakely JA, Easton JD, et al. The Canadian American Ticlopidine Study in Thromboembolic Stroke. Lancet 1989; i: 1215-20. 6. Bamford J, Sandercock P, Dennis M, et al. A prospective study of acute cerebrovascular disease in the community: the Oxfordshire Community Stroke Project 1981-86. 1. Methodology, demography and incident cases of first-ever stroke. J Neurol Neurosurg Psychiatry 1988; 51: 1373-80. 7. Bamford J, Sandercock P, Dennis M, Burn J, Warlow CP. A prospective study of acute cerebrovascular disease in the community: the Oxfordshire Community Stroke Project 1981-86. 2. Incidence, case fatality rates and overall outcome at one year of cerebral infarction, primary intracerebral haemorrhage and subarachnoid haemorrhage. J Neurol Neurosurg Psychiatry 1990; 53: 16-22. 8. Hatano S. Experience from a multicentre stroke register: a preliminary report. Bull WHO 1976; 54: 541-53. 9. Sandercock P, Allen C, Corston R, Harrison M, Warlow C. Clinical

1526

diagnosis of intracranial haemorrhage using Guys Hospital Score. Br Med J 1985; 291: 1675-78. 10. Bamford JM, Warlow CP. The evolution and testing of the lacunar hypothesis. Stroke 1988; 19: 1074-82. 11. Hommel M, Besson G, Le Bas JF, et al. Prospective study of lacunar infarction using magnetic resonance imaging. Stroke 1990; 21:
546-54. 12. Bamford JM. Non ischaemic causes of lacunar syndromes: prevalence and clinical findings. J Neurol Neurosurg Psychiatry 1990; 53: 819. 13. Fisher CM. Capsular infarcts&mdash;the underlying vascular lesions. Arch Neurol 1979; 36: 65-73. 14. Allen CMC, Hoare RD, Fowler CJ, Harrison MJG. Clinico-anatomical correlations of uncomplicated stroke. J Neurol Neurosurg Psychiatry 1984; 47: 1251-54. 15. Kase CS. Middle cerebral artery syndromes. In: Vinken PJ, Bruyn GW, Klawans HL, Toole JF, eds. Handbook of clinical neurology. Amsterdam: Elsevier, 1988: 353-70. 16. Escourelle R, Poirier J. Manual of basic neuropathology. Philadelphia: WB Saunders, 1978. 17. Bladin PF, Berkovic SF. Striato-capsular infarction: large infarcts in the lenticulostriate arterial territory. Neurology 1984; 34: 1423-30. 18. Caplan LR. Vertebrobasilar system syndromes. In: Vinken PJ, Bruyn GW, Klawans HL, Toole JF, eds. Handbook of clinical neurology. Amsterdam: Elsevier, 1988: 353-70. 19. Caplan LR. "Top of the basilar syndrome". Neurology 1980; 31: 72-79. 20. Bamford J, Sandercock P, Warlow C. Interobserver agreement for the assessment of handicap in stroke patients. Stroke 1989; 20: 828.

21. Bamford J, Dennis M, Sandercock P, Burn J, Warlow C. The frequency, causes and timing of death within 30 days of a first stroke: the

Community Stroke Project. J Neurol Neurosurg Psychiatry 1990; 53: 824-29. 22. Pickard JD, Murray GD, Illingworth R, et al. Effect of oral nimodipine on cerebral infarction and outcome after subarachnoid haemorrhage: British aneurysm nimodipine trial. Br Med J 1990; 298: 636-42. 23. Hart RG. Prevention of cardioembolic stroke. In: Furlan AJ, ed. The heart and stroke. London: Springer-Verlag, 1987: 117-37. 24. Humphrey PRD, Harrison MJG. How often can embolic stroke be diagnosed clinically? A clinico-pathological correlation. Postgrad Med J 1985; 61: 1039-42. 25. Whisnant JP, Basford JR, Bernstein EF, et al. Classification of
Oxfordshire
cerebrovascular diseases III. Stroke 1990; 21: 637-76. 26. Harrison MJG. Clinical distinction of cerebral haemorrhage and cerebral infarction. Postgrad Med J 1980; 56: 629-32. 27. Foulkes MA, Wolf PA, Price TR, Mohr JP, Hier DB. The Stroke Data Bank: design, methods and baseline characteristics. Stroke 1988; 19: 547-54. 28. Brott T, Mohr JP, Olinger CP, et al. Measurement of acute cerebral infarction: lesion size by computed tomography. Stroke 1989; 20: 871-75. 29. Chodosh EH, Foulkes MA, Kase CS, et al. Silent strokes in the NINCDS Stroke Data Bank. Neurology 1988; 38: 1674-79. 30. Turney TM, Garraway WM, Whisnant JP. The natural history of hemispheric and brainstem infarction in Rochester, Minnesota. Stroke

1984; 15: 790-94.

SCREENING
Case-control study of screening for prostatic by digital rectal examinations
cancer

Although commonly thought to be an effective method of screening for prostatic cancer, digital rectal examinations have yet to be shown by controlled study to help prevent advanced forms of this disease. 139 men with metastatic (stage D) prostatic cancer (cases) were compared with an equal number of matched men free of this condition (controls), with respect to rectal examinations
recorded in the medical records up to, on average, 23 years before the cases initial diagnosis of prostatic cancer. Cases and controls were members of a large health maintenance organisation in northern California. In the 10 years before initial diagnosis (excluding the last three months) the average number of examinations for routine screening (2&middot;45 vs 2&middot;52) or to evaluate intestinal or rectal symptoms (0&middot;44 in both) were similar in cases and controls, respectively. After adjustment for racial differences, the relative risk of metastatic prostatic cancer for men with one or more screening rectal examinations compared with men with none was 0&middot;9 with a 95% confidence interval of 0&middot;5-1 &middot;7. Screening by routine digital rectal examination appears to have little if any effect in preventing metastatic prostatic cancer. If there is a small benefit, it will be difficult to demonstrate by conventional epidemiological study.

Introduction
One purpose in including digital rectal examination in the routine health examination of men is to detect asymptomatic cancer of the prostate. The hope is that early detection can lead to therapy that will cure this disease, or at least prevent or delay its spread beyond the prostate gland. Rectal examinations have low sensitivity and specificity in the detection of prostatic cancer1 and their value in screening is controversial.2-5 We know of no controlled studies of whether rectal examinations do help to prevent metastases of prostatic cancer. We have done a case-control study comparing, in men with metastatic carcinoma of the prostate and in similar men free of this condition, history of rectal examinations before prostatic cancer was diagnosed. If screening rectal examinations were of value, one would expect to find substantially fewer of them in the cases.

Methods
The Kaiser Permanente Medical Care Program, Northern California Region, is a large health maintenance organisation serving residents of the San Francisco Bay area and nearby communities. Subscribers comprise about one-fourth of the population residing in the areas served and are heterogeneous, both
ADDRESS Division of Research, Kaiser Permanente Medical Care Program, Northern California Region, Oakland, California, USA (G. D. Friedman, MD, R. A. Hiatt, MD, C P. Quesenberry, Jr, PhD, J. V Selby, MD) Correspondence to Dr Gary D Friedman, 3451 Piedmont Avenue, Oakland, California 94611-5400, USA.