Preventative of brain injury (cerebral protection) is only possible in certain circumstances such as in patients undergoing cardiopulmonary bypass. Hypothermia and the barbiturates are particularly effective for this. Therapy commenced during a cerebral insult may be referred to as cerebral preservation while cerebral resuscitation describes any therapy begun after brain injury. Cerebral preservation occurs in the cold water drowning victim who makes a good neurological recovery despite a prolonged period beneath the water.

Several processes interact at cellular level to determine the viability of brain tissue following injury. Each of these processes can be altered or ameliorated by brain-orientated intensive care. The rapid re-establishment of cerebral blood flow is of critical importance: brain oxygen stores are depleted in about 10 s and brain glucose and ATP stores are depleted 5 min following circulatory arrest. Some cerebral neurones are able to tolerate normothermic ischaemia for somewhat longer periods and the most extreme form of cellular damage, autolysis of brain tissue, begins after 1 to 2 h of no blood flow.

After circulatory arrest of more than 5 min, reperfusion and reoxygenation may cause irreversible brain damage, in the same way that other organ systems may be damaged. However, the widely held belief that a normothermic circulatory arrest lasting longer than 5 min is incompatible with recovery of normal brain function is not always correct. Several studies in animal models have shown good cerebral recovery after periods of circulatory arrest of more than 15 min and there have been reports of humans recovering after arrest times of up to 15 min.

The degree of brain injury, regardless of the nature of the insult, depends first upon the severity and duration of the injury and second, upon the speed and efficiency of resuscitation. To this should be added a third factor, the early application of brain-orientated intensive care.

Brain-orientated intensive care is broadly based and comprises several modalities (Table 1) 73. All organ systems must be actively supported and function restored as quickly as possible. The period immediately following cardiopulmonary resuscitation for circulatory arrest may be characterized by persistent metabolic acidosis and impaired cardiac contractility: the myocardium may need to be supported with inotropes and gas exchange maintained by mechanical ventilation until any acidosis resolves. Bicarbonate infusion is best avoided since alkalinizing agents may worsen cerebral and myocardial intracellular acidosis.

The systemic blood pressure should be supported with vasopressors if autoregulatory mechanisms are impaired. Cardiac arrhythmias must be rigorously controlled. Biochemical abnormalities such as hyperglycaemia should be corrected. If sodium bicarbonate has been used injudiciously during resuscitation, serum osmolality may be significantly increased: the use of mannitol as an osmotic diuretic could further exaggerate the hyperosmolar state and serial measurements of serum osmolality may need to be made. Hyperglycaemia should be controlled with insulin if necessary since preischaemic hyperglycaemia has been shown to worsen the outcome of both global and focal cerebral ischaemia.

The management of blood pressure in patients with brain injury and raised intracranial pressure is controversial. Hypotension carries the risk of global hypoperfusion, while hypertension, although maintaining perfusion pressure, encourages the development of cerebral oedema. Common sense dictates a strategy aimed at avoiding extremes of blood pressure, with target systolic blood pressures between 100 and 160 mmHg. To this end, direct intra-arterial blood pressure monitoring is preferred to non-invasive methods. Theoretically, vasodilators such as hydralazine, nitroprusside, and nitroglycerin should be avoided since they could cause stealing of blood flow away from the brain. In practice, however, agents such as nitroprusside and nitroglycerin are very effective in reducing blood pressure.

Hyperthermia unrelated to infection may be observed in the early phases following head injury or brain ischaemia. Active efforts to lower the temperature should be made with antipyretic (acetaminophen) and cooling. Small intravenous doses of chlorpromazine (2.5 mg for adults) may be effective when other measures fail. It is well recognized that hypothermia at the time of cerebral injury protects the brain against anoxic insult.

Cerebral oedema and the reduction in cerebral perfusion brought about by raised intracranial pressure is a particular problem in trauma and encephalitis. Almost 80 per cent of patients with severe head injury have some degree of elevation in intracranial pressure; 50 per cent of deaths are directly attributable to raised intracranial pressure. It has also been estimated that up to one-third of deaths due to head injury may occur without any such increase. Cerebral oedema following circulatory arrest is less commonly associated with increased intracranial pressure. The normal pressure, of less than 15 mmHg at rest, with wide swings according to body activity and position, is determined by the relationship of the intracranial and intraspinal volume to the restricted rigid skull vault and the more compliant spinal subarachnoid space. The volume compartments within the skull are the brain, the cerebrospinal fluid, and the intravascular blood. Cerebrospinal fluid and blood are displaced as brain tissue swells. The relationship between intracranial pressure and increases in volume is not linear, but is roughly exponential. At high pressures (>25 mmHg) a small increase in volume causes a steep rise in pressure, but little damage is caused until cerebral blood flow becomes restricted. An intracranial pressure which is sustained at above 50 mmHg results in greatly reduced cerebral perfusion and is usually fatal.

Considerable controversy exists regarding the effects on outcome of monitoring intracranial pressure. Evidence to show that such monitoring significantly improves survival following head trauma is less than convincing, but neurological deterioration often appears to follow episodes of raised intracranial pressure, and patients with persistently raised intracranial pressure generally have a poor outcome. If attempts to lower intracranial pressure are to be made at all, measurement would seem logical: the clinician can then titrate therapy to achieve an acceptable cerebral perfusion pressure (mean arterial pressure minus mean intracranial pressure) of greater than the critical level of 40 mmHg. Perfusion pressures in the range 60 to 80 mmHg are ideal, but are not always attainable.

Clinical signs are poor indicators of intracranial pressure: temporal lobe lesions may be associated with third nerve signs at a lower pressure than lesions elsewhere and diffuse bilateral cerebral lesions cause pupillary enlargement only when pressure is considerably greater than 25 mmHg. Sixth nerve palsy is commonly described but is not a reliable indicator of raised intracranial pressure.

Currently available techniques for the measurement of intracranial pressure use either a hollow subarachnoid screw (Richmond bolt), ventricular catheter, or miniature strain gauges. Intraventricular catheters probably carry a higher risk of infection but also provide a more reliable measurement.

A recent innovation, infrared transmission cerebral spectroscopy, may provide a convenient and non-invasive method of evaluating the state of oxygenation of the brain by measuring the haemoglobin saturation of predominantly mixed venous cerebral blood. Whether this technique will provide clinically useful information remains to be seen but initial evaluations appear encouraging.

Mechanical hyperventilation to reduce the Pco&sub2; to below 3.5 kPa (<30 mmHg) will rapidly lower intracranial pressure by causing cerebral vasoconstriction. These effects are seen within minutes, but are dissipated within 1 to 2 h. Further reduction in intracranial pressure requires a further increase in the level of hyperventilation and reducing the level of ventilation may result in an increase in pressure unless the underling pathology has been reversed. An excessive reduction in Pco&sub2; or elevation in arterial pH may reduce cerebral perfusion and oxygen off-loading and may therefore be counter productive. Hyperoxia should probably be avoided. Hyperventilation necessitates higher mean airway pressures and the ventilator should be regulated to produce the highest alveolar ventilation for the lowest inflation pressures. This can often be attained using an assisted mode such as volume assist or pressure assist in a non-paralysed patient. If the patient is intolerant of this approach and becomes agitated, full sedation and, if necessary, neuromuscular relaxation will allow the patient to be maintained on SIMV (synchronized intermittent mandatory ventilation) or a control mode of ventilation. Although hyperventilation is an established mode of treatment of raised intracranial pressure, there are no clinical trials proving its efficacy, and one recent study has suggested poorer long-term outcome (3–6 months) in patients hyperventilated following head injury. In experimental models of brain injury, hyperventilation neither reduces brain lactate nor improves the energy state (phosphocreatine: inorganic phosphate ratio) of the brain.

Hyperosmolar therapy with mannitol and frusemide (furosemide) is widely recommended for production of a sustained reduction in intracranial pressure. Mannitol can be administered as a single bolus dose of 0.5 to 1 g/kg body weight, or repeated as an infusion of 0.25 to 0.5 g/kg. This induces an osmotic gradient that reduces intracranial pressure for 3 to 4 h in most patients. Side-effects, due to systemic dehydration and intravascular volume depletion, including the precipitation of a non-ketotic hyperosmolar state in diabetics, may be seen with mannitol. Frusemide (furosemide) complements the action of mannitol, but may be less effective when used alone.

Barbiturate administration following acute cerebral injury may have several beneficial effects. Control of agitation and restlessness by sedative agents minimizes intracranial pressure changes, and seizures should be prevented. Barbiturates may also possess specific cerebral protective properties, although this is contentious and offers an avenue for further investigation. Clinical studies in humans have produced conflicting results, but some cerebral protection appears to have accrued from the use of barbiturates in patients with Reye's syndrome. There are sound reasons why barbiturates should protect the brain following injury (Table 2) 74.

Of the barbiturates available, thiopentone (pentothal) is an excellent sedative agent and anticonvulsant. With prolonged use it is distributed widely throughout the body and its effects are then slow to clear. Until properly conducted clinical studies can convincingly show a specific cerebral protective effect of barbiturates, justification for their use will be limited to the sedative and anticonvulsant effects.

The calcium-channel blockers, particularly nimodipine, improve survival following subarachnoid haemorrhage by reversing cerebral vasospasm. They probably act by reducing the intracellular release and accumulation of free ionized calcium and inhibition of mitochondrial activity, which occurs during ischaemia and reperfusion. Studies in animals and in man have examined the potential protective effect of calcium-channel blockers in a much wider range of cerebral injury with some indication of a beneficial response following cardiac arrest. Despite a degree of cerebral selectivity, the use of calcium-channel blockers may be limited by adverse effects such as systemic hypotension.

Free radicals are short-lived highly reactive compounds that are released during the reperfusion of ischaemic neuronal issue and initiate sustained lipid peroxidation. Free radical scavengers such as superoxide dismutase, desferrioxamine, thiopentone, vitamin E, vitamin C, glutathione, chlorpromazine, mannitol, and some dextrans have been used in an attempt to reduce the effects of reperfusion injury of several organs including the brain. Preliminary studies have shown no protective effect of desferrioxamine and a calcium-channel blocker in a dog ischaemic brain model, although desferrioxamine alone appeared to reduce cerebral damage following shorter periods of cerebral ischaemia. The use of free radical scavengers cannot yet be recommended, however rational their application may seem.

Corticosteroids have no beneficial effect in cerebral oedema associated with head injury, and their effects following cardiac arrest have not been adequately explored. Their use in brain-orientated intensive care, other than to reduce high intracranial pressure associated with brain metastases, cannot be recommended.

The interaction of certain drugs with &ggr;-aminobutyric acid receptors may have some application in cerebral protection. Drugs such as diazepam (benzodiazepine receptor agonist), baclofen (&ggr;-aminobutyric acid B agonist), valproic acid (&ggr;-aminobutyric acid transaminase inhibitor), and pentobarbital (an effector of &ggr;-aminobutyric acid A receptors) require further evaluation as protective agents.

Coma is a state of depressed level of consciousness and should be distinguished from the abnormal content of consciousness that results in confusion, delirium, or psychosis. The drowsy patient (level) may also be confused (content), although the converse may not be true. The level of consciousness should always be quantified in terms of some standard such as the Glasgow Coma Score (Table 3) 75. By accurately recording coma scores, small changes in conscious level can be recognized and if necessary, acted upon. Most patients admitted to intensive care in coma will have a score of less than 6.

The level of consciousness emanates from the reticular activating system and the cerebral hemispheres. Transient loss of consciousness (concussion) after head trauma has been attributed to rotation of the hemispheres about the midbrain/diencephalic junction. Ultrastructural changes occur in the brain neurones, even with very transient loss of consciousness after trauma. Prolonged loss of consciousness (coma), whether or not related to trauma, is usually associated with lesions in either both hemispheres, one hemisphere with compression of the upper brain-stem, or in the brain-stem reticular activating system. When coma is due to injury to the reticular activating system, other signs of brain-stem dysfunction such as pupillary, caloric, and oculomotor signs are usually apparent.

The predominant causes of coma vary from centre to centre: drug intoxication and head injury may be more common in the inner city while metabolic disease such as diabetic coma, cerebrovascular events, and mass lesions may predominate in other hospitals. A careful clinical history and physical examination are therefore paramount in determining the underlying cause of coma. A list of common causes of coma is shown in Table 4 76.

Diagnostic dilemmas often arise when coma is due to drugs or metabolic causes, when physical signs may be consistent with both cerebellar and brain-stem lesions, depending upon the severity of the metabolic disturbance. However, pupil size and reactivity are usually well preserved despite severe obtundation.

Patients with severe myasthenia gravis, Guillain-Barré syndrome, or basilar artery thrombosis may appear to be in coma although they may be receptive. The so called ‘locked-in’ syndrome due to cranial nerve and limb paralysis may be extremely difficult to distinguish from coma: the clinical significance of this becomes only too apparent when considering patients as potential organ donors.

Computerized tomography (CT) of the head may define anatomical abnormalities and confirm a clinical diagnosis. Although the CT scan is generally a safe procedure and is available to most critically ill patients it should not be used indiscriminantly, bearing in mind the risks associated with transporting the patient to the radiography department. Conversely, changes in coma score may mean that repeated scans are necessary to determine the requirement for surgical intervention.

The pupillary reflex must be assessed with care and attention: a cursory evaluation with a poor light source is clearly inadequate. Contraction of only 1 mm is sufficient to indicate pupillary responsiveness. Unequal pupils may be found following the instillation of mydriatics or direct trauma to the eye.

The presence of full and conjugate eye movements indicates an intact pons and midbrain. Normally, turning the head from side to side will cause the eyes initially to move conjugately in the opposite direction followed quickly by eye movement in the direction of head movement. The presence of ‘doll's eyes’, which remain fixed in relation to the head, indicates brain-stem damage and should be confirmed by caloric testing. Abducted (outward looking) eyes are common in stuporous or drowsy patients and should not be interpreted as a bad prognostic sign. As coma deepens the eyes may become fixed in the primary position.

Limb movement affords the most reliable method of detecting asymmetric neurological function. Movements of a limb away from the body (i.e. pushing away) indicate an intact corticospinal pathway and this type of purposeful limb should always be distinguished from flexion, extension, or pronation. The triple flexion response of the hip, knee, and ankle joints in response to tactile stimuli is not necessarily purposeful since it is integrated at the level of the spinal cord.

Protection of the airway takes precedence over all other aspects of management. Failure to protect the airway and control the cervical spine in a comatose patient with head injury must be considered to be serious omissions. Positioning of the patient in the semiprone or Fowler's position may be adequate in the drowsy patient with an intact gag reflex but as consciousness fades, endotracheal intubation may be required. The technique of intubation should combine speed, safety, and avoidance of factors that may increase intracranial pressure.

Predicting the outcome of coma, regardless of the underlying cause, is a difficult task. Apart from unequivocal brain death there are no clinical signs that confidently predict outcome. Young patients may have many early clinical signs consistent with poor outlook and yet may make a full recovery. Assessment of prognosis based on neurological findings should be interpreted broadly, and in conjunction with the patient's age and prior medical condition. The clinician's approach to management should also incorporate a consideration of the patient's premorbid wishes, if these are known.

Prognostication in cases of head injury is usually more accurate than in non-traumatic coma, for which there are many aetiologies. While more than 90 per cent of patients in whom pupillary or oculomotor reflexes are absent in the first 6 h following injury will die, 4 per cent may still make a significant recovery: this proportion is large enough to make the clinician hesitate to abandon support in the first 24 h. Some favourable and unfavourable signs of recovery in the comatose head injury patient are summarized in Table 5 77. Such signs can be cautiously applied to patients with coma due to other causes, but with less confidence.

Somatosensory evoked potentials may offer useful information, particularly in children. Absence of cortical responses with preserved lower level brain potentials appears to be a reliable indicator of poor outlook in anoxic coma and following head injury.

In practice, an expectant policy allowing time for the development of convincing signs, full evaluation, and neurological consultation to take place seems to be the most acceptable. In those patients in whom the outlook is grave, family and relatives can come to terms with reality. A period of 48 to 72 h may be needed initially. The primary physician should indicate clearly the management policy to nursing staff and house staff, particularly after full and frank discussion with the nearest relatives. In some cases, extubation may be possible and more conservative management plans adopted.

The accurate and reliable determination of brain death has become an essential part of clinical practice in the critical care unit. Brain-dead patients need to be identified so that either unnecessary life support can be discontinued or so that transplantation organ donation can be considered and organized. It is unnecessary to establish the diagnosis of brain death, unless organ transplantation or the cessation of life support measures is deemed appropriate.

The criteria used acknowledge the fact that independent, self-sustaining life is not possible in patients with brain-stem death. It is also inevitable that within a short period of time (usually hours, but at most 2 or 3 days) circulatory collapse and cardiac arrest will occur which will be unresponsive to any medical measures.

Before undertaking brain death testing in the comatose patient several preconditions must be satisfied. First, a definitive cause for coma must be established. Second the patient requires mechanical ventilation and third, enough time must have elapsed from the onset of coma to determine that the brain injury is irreversible. This time interval will vary according to the cause of the brain injury. For example, following head trauma or a major intracranial haemorrhage 6 to 12 h may be sufficient time to be sure that recovery is not possible. In the case of hypoxic brain injury secondary to cardiac or respiratory arrest 24 to 48 h may be needed. If any doubt exists as to possible effects of drugs, periods extending up to 7 days may need to be considered. The elimination of drugs and duration of activity is extremely variable. Some agents, such as nortriptyline, diazepam, methadone, and phenobarbitone have prolonged plasma half-lives of between 10 and 150 h and may therefore have very extended activity.

The tests of brain-stem function include many of the cranial nerve reflexes and the activity of the respiratory centre. Pupillary reaction to light depends upon the integrity of the optic nerve, the midbrain, and, the oculomotor nerve (III). The corneal reflexes are dependent on the afferent trigeminal (V) and the facial nerve (VIII).

Dysfunction of both cerebral hemispheres disinhibits the brainstem reflex mechanisms for conjugate eye movements. Normally, turning the head from side to side elicits easy or ‘loose’ conjugate eye movements in the oppositive direction. Brain-stem dysfunction cause one or both eyes to fail to move fully and conjugately. Abducens nerve (VI) palsy indicates pontine dysfunction or diffuse increase in intracranial pressure. Oculomotor nerve (III) palsy with absence of full adduction of one eye also indicates midbrain pathology. The pattern of spontaneous respiration may become periodic when both hemispheres are affected, and irregular or absent when the inferior pons and medulla are involved.

Established hospital guidelines for the diagnosis of brain death should be rigorously adhered to. The completion of a standard form to be made part of the patient's record is generally desired. Table 6 78 lists many of the most widely accepted brain death criteria.

Specific criteria for the diagnosis of brain death may vary in minor details depending on local or national recommendations. These variations may include the requirement for two sets of observations, about 24 h apart, and an EEG showing electrocerebral silence. An advantage of performing two tests lies not in improved accuracy, but provides time for the co-ordination of transplant teams. Some protocols require the radiographic demonstration of absent cerebral blood flow.

The conduct of the apnoea test may also vary from centre to centre but relies on achieving an indisputable CO&sub2; stimulus to respiration in the absence of respiratory depressants such as sedatives, narcotics, neuromuscular relaxants, or hypoxia. The patient should be monitored with continuous ECG and pulse oximetry if available. Mechanical ventilation should be adjusted to stabilize the Pco&sub2; at a normal high level. The patient should then be pre-oxygenated with 100 per cent O&sub2; for about 10 min. The ventilator can then be disconnected at the endotracheal tube and a catheter passed down the endotracheal tube to facilitate insufflation of oxygen at 1 to 2 min. The patient is carefully observed to detect any respiratory movement of the chest or abdomen. Arterial oxygen saturation can be accurately monitored with a pulse oximeter. After 5 to 10 min blood gases can again be sampled. The patient is then temporarily supported with the ventilator until the blood gases confirm a high CO&sub2; stimulus (6.5 kPa, 50 mmHg) and the absence of hypoxia. In practice Pco&sub2; values between 8 and 10 kPa can be achieved by this protocol. If no respiratory effort has been detected and other signs of brain-stem dysfunction have been demonstrated the patient may be considered brain dead.

It is the responsibility of the doctor performing the tests of brain death to explain the implications of such tests to relatives before the tests are completed. It is essential that the relatives are aware that survival is not possible in the presence of positive brain death criteria.

Abramson NS. et al. Randomized clinical study of thiopental loading in comatose survivors of cardiac arrest. N Engl J Med, 1986; 314: 397–403.

Ames A III, Nesbett FB. Pathophysiology of ischemic cell death. I. Time of onset of irreversible damage; importance of the different components of the ischemic insult. Stroke, 1983; 14: 219–23.

Bircher N, Safari P. Cerebral preservation during cardiopulmonary resuscitation. Crit Care Med, 1985; 13: 185–90.

Bruce DA, Gennarelli T, Langfitt T. Resuscitation from coma due to head injury, 1978; Crit Care Med, 6: 254–8.

Cooper P. et al. Dexamethasone and severe head injury. A prospective, double blind study. J Neurosurg, 1979; 51: 307–10.

Farber JL, Chien KR, Mittnacht S. The pathogenesis of irreversible cell injury in ischemia. Am J Physiol, 1981; 102: 271–81.

Fleischer JE, et al. Failure of deferoxamine, an iron chelator, to improve neurologic outcome following complete cerebral ischemia in dogs. Stoke, 1987; 18: 124–7.

Gudeman S, Miller J, Becker D. Failure of high-dose steroid therapy to influence intracranial pressure in patients with severe head injury. J Neurosurg, 1979; 51: 301–6.

Jennet B, et al. Predicting outcome in individual patients after severe head injury. Neurosurgery, 1979; 4: 283–8.

Kirsch JR, Helfaer MA, Koehler RC, Traystman RJ. Brain ischaemia and reperfusion injury. In Bihari D, Holiday JW, eds. Update in Intensive Care and Emergency Medicine. Berlin: Springer-Verlag, 1989; 66–84.

Lim KH. et al. Prevention of reperfusion injury of the ischemic spinal cord: use of recombinant superoxide dismutase. Ann Thoracic Surg, 1986; 42: 282–6.

Lind B, Snyder J, Safer P. Total brain ischemia in dogs: cerebral physiological and metabolic changes after 15 minutes of circulatory arrest. Resuscitation, 1975; 4: 97–101.

Marshall LF, et al. Pentobarbital therapy for intracranial hypertension in metabolic coma. Reye's syndrome. Crit Care Med, 1978; 6: 1–6.

McCormick PW, et al. Noninvasive cerebral optical cerebral spectroscopy for monitoring cerebral oxygen delivery and haemodynamics. Crit Care Med, 1991; 19: 89–97.

Michael R, et al. Mechanisms by which epinephrine augments cerebral and myocardial perfusion during cardiopulmonary resuscitation in dogs. Circulation, 1984; 69: 822–35.

Pallis C. ABC of brain stem death. Torquay: British Medical Journal/Devonshire Press, 1983.

Plum F, Posner J. The Diagnosis of Stupor and Coma. Philadelphia: Davis, 1980.

President's Commission for the Study of Ethical Problems in Medicine and Biomedical and Behavioral Research. Defining Death. Medical, Legal and Ethical Issues in the Determination of Death. Washington: US Government Printing Office, 1981.

Rockoff M, Marshall L, Shapiro H. High-dose barbiturate in therapy humans: A clinical study of 60 patients. Ann Neurol, 1979; 6: 194–8.

Safar P, Bircher N. Cardiopulmonary cerebral resuscitation. An Introduction to Resuscitation Medicine. World Federation of Societies of Anaesthesiologists. 3rd edn. London: Balliere Tindall, 1986.

Safar P, Grenvik A, Abramson NS, Burcher N, eds. Reversibility of clinical death: symposium on resuscitation research. Crit Care Med, 1988, 16: 919–1086.

Schanne FAX, et al. Calcium dependence of toxic cell death: A final common pathway. Science, 1979; 206: 700–3.

Shapiro HM. Postcardiac arrest therapy. Calcium entry blockade and brain resuscitation. Anesthesiology, 1985; 62: 384–7.

Shiu GK, Nemoto EM. Barbiturate attenuation of brain free fatty acid liberation during global ischemia. Neurochemistry, 1981; 37: 1448–56.

Smith DS, Rehncrona S, Siesjo BK. Barbiturates as protective agents in brain ischemia and as free radical scavengers in vitro. Acta Physiol Scand, 1980; 42: 129–34.

Steen PA, et al. Cerebral blood flow and neurologic outcome when nimodipine is given after complete cerebral ischaemia in the dog. J Cerebral Blood Flow Metab, 1984; 4: 82–7.

Todd MM, Chadwick HS, Shapiro HM, Dunlop BS, Marshall LF, Dueck R. The neurologic effects of thiopental therapy following experimental cardiac arrest in cats. Anesthesiology, 1982; 57: 76–86.

Wei EP, et al. Effects of oxygen Radical on cerebral arterioles. Am J Physiol, 1985; 248: H157–H162.