Controversies in Management: Computed tomography provides accurate diagnosisBMJ 1994; 309 doi: https://doi.org/10.1136/bmj.309.6967.1498 (Published 03 December 1994) Cite this as: BMJ 1994;309:1498
- Joanna M Wardlaw, senior registrar in neuroradiologya
The term stroke describes a clinical event that can be due to several underlying conditions. Patients with stroke are investigated to help clinicians to separate strokes from non-strokes; distinguish cerebral haemorrhage from cerebral infarction; and identify specific pathophysiological subtypes of cerebral infarction. This information aids decisions about treatment and helps determine prognosis.1 Investigations are of clinical benefit only if they improve diagnostic accuracy beyond that which can be obtained from the history and examination of the patient. Computed tomography of the brain is the quickest, most practical method of diagnosing the cause of stroke.
What do “experts” think?
Several statements from experts have suggested that computed tomography is useful in patients with stroke. In 1988 the King's Fund issued a consensus statement that said: “There is a strong case for scanning most patients who present with a presumptive stroke, excluding those in whom antithrombotic treatment is contraindicated. The resources required to scan this large group of patients may well be outweighed by a reduction in the recurrence of strokes and myocardial infarcts and in the associated costs of health care.”2 In 1990 the Royal College of Physicians of London concluded that “All patients who present with the acute onset of a focal neurological deficit should have a CT scan.”3 In 1993 the United States National Stroke Association issued a consensus statement: “CT has become absolutely necessary for patients suspected of having a stroke because all subsequent therapeutic decisions depend on its results.”4 But are all these “experts” necessarily correct? What is the hard evidence that computed tomography is of benefit in patients with stroke?
Is computed tomography useful?
In 1985 the Oxfordshire Community Stroke Project examined the usefulness of computed tomography. The project found that computed tomography provided useful information in a minority (up to 28%) of patients in whom the diagnosis of stroke was uncertain, antithrombotic treatment was being given or contemplated, carotid endarterectomy was contemplated, cerebellar haematoma was suspected, or deterioration was atypical of stroke. However, important advances have been made in primary and secondary prevention of stroke since 1985.
Aspirin,6 carotid endarterectomy,7 and anticoagulation8 are of proved benefit in appropriately selected patients. More is known about the likely cause, the prognosis, and risk of recurrence of different clinical subtypes of cerebral infarction.9 Investigation and treatment of stroke patients can therefore be targeted to ensure cost effective use of resources.9 Despite these advances in prevention of stroke no effective primary treatment exists for cerebral infarction. Clearly, if a primary treatment for acute ischaemic stroke became available, computed tomography would become even more important.
Clinical diagnosis of stroke: how good is it?
Outside hospital about 80% of strokes are cerebral infarcts, 10% are primary intracerebral haemorrhages, 5% are subarachnoid haemorrhages, and about 5% are of uncertain cause.10 In hospital only 60-75% of patients with stroke have a cerebral infarct. If doctors simply assumed that most patients with stroke admitted to hospital had had cerebral infarction, the diagnosis could be wrong in 35-40% of patients.
In the Oxford study 18% of patients referred to the investigators by physicians were considered not to have had a stroke clinically when seen by the neurologist.5 In the pilot phase of the international stroke trial of antithrombotic treatment in acute ischaemic stroke 43 (16%) of 265 patients referred by hospital doctors to the trial were considered not to have had a stroke clinically when seen by the trial physician (R I Lindley, personal communication).
Clinical scoring methods to distinguish haemorrhage from infarct such as the Allen11 or Siriraj12 scores increase clinical diagnostic accuracy, but two prospective studies have shown that even in patients with a low probability (less than 10%) of cerebral haemorrhage on the basis of these scores, at least 7% (with the Allen) and 5% (with the Siriraj) of patients had an intracerebral haemorrhage (S Ricci, personal communication; R I Lindley et al, second international conference on stroke, Geneva, May 1993). Thus even with a complex and rather time consuming clinical scoring method and even in patients with the lowest probability of haemorrhage the clinical differentiation of cerebral infarction from haemorrhage will still be wrong, importantly wrong, in up to 10% of patients. Furthermore, computed tomography must be done within seven to 10 days after stroke otherwise the radiological signs of haemorrhage may have evolved, leaving an appearance identical to an infarct.
Arguments against routine computed tomography
If computed tomography is not used routinely, which stroke patients should have the benefit of correct diagnosis? Should it just be “young” patients, and if so what would be the upper age limit? The average life expectancy of an 87 year old woman in the United States is 5.6 years. The average life expectancy in the United Kingdom may not be as long but that does not mean that the chance of independent life should be denied to elderly people for the sake of a single investigation.
The cost and limited availability of computed tomography are not valid reasons for not scanning patients. In Britain a brain computed tomogram (about £100) costs less than an average day in hospital (£200-£300). Timely use of computed tomography to guide patient management could shorten hospital stays and allow more appropriate use of ancillary investigations. For example, indiscriminate use of aspirin in presumed ischaemic stroke, inappropriate referrals for carotid endarterectomy, or inappropriate use of antithrombotic drugs (stopping or starting)—resulting in even as few as 10 extra strokes per 100000 population, with an average bed occupancy of 30 days, would easily be equivalent to the cost of scanning all stroke patients. It doesn't make sense not to scan stroke patients when the balance of risk and benefit in medicine is so tight.
In 1993 the Stroke Association surveyed all British physicians.14 They were asked if they would want computed tomography if they had a stroke; 90% said “Yes.” If the people with most experience of caring for stroke patients would like to have a computed tomography if they had a stroke themselves, why should patients be denied this facility? The evidence for routine use of computed tomography in acute stroke is overwhelming.