In the United States, subarachnoid hemorrhage (SAH) from intracranial aneurysms occurs with an annual incidence of 12 per 100,000 (1) and has a one-month mortality rate approaching 50%(2). The leading causes of death and morbidity after SAH are the immediate effect of SAH itself, rebleeding, and cerebral ischemia and infarction from cerebral vasospasm (2). In humans, aneurysmal vasospasm is believed to be biphasic, with an immediate acute phase within hours and a delayed phase typically beginning two to four days after SAH and peaking on approximately Day 7 (3,4). Whereas vasospasm has been relatively easy to demonstrate in large conduit cerebral arteries, it has been questioned whether cerebral microvessels are also prone to vasospasm (5,6).
In the present study, we used a rat model of SAH, in which the acute phase of vasospasm was consistently demonstrated in large cerebral arteries (7). We examined in vitro whether the cerebral cortical microvessels demonstrated altered vasomotor responses. We tested the hypothesis that after SAH, microvascular endothelial dysfunction occurs that may predispose the microvessels to spasm. Further, we examined whether the observed endothelial dysfunction may be related to altered expression of endothelial nitric oxide synthase (NOS3).
With institutional animal care and use committee approval, Wistar rats of either sex, weighing 250–300 g, were anesthetized by injecting 40 mg/kg ketamine and 5 mg/kg xylazine intraperitoneally. A polyethylene catheter was inserted percutaneously into the cisterna magna (CM) and was secured in place. After 2 days of recovery, lack of gross wound infection was noted, and normal neurological function was verified by a lack of forelimb flexion on tail suspension, a lack of lateral sliding, and a lack of circling behavior (8). The animal was then reanesthetized with ketamine and xylazine as before. A femoral arterial catheter was placed for blood sampling. SAH was modeled by withdrawing 0.3 mL of cerebrospinal fluid from the CM and injecting 0.3 mL of fresh nonheparinized autologous blood into the CM slowly for a 10-min period with the animal in a 20° head-down position, as described by Cole et al. (7). The animal was allowed to breathe spontaneously and kept in the head-down position. Twenty minutes after the subarachnoid injection of blood, at a time known to coincide with maximal vasospasm (7), a femoral arterial blood was sampled for blood gas analysis, and the animal was killed for the harvest of brain tissue, which was quickly placed in a cold (4°C) modified Krebs solution (120 mM NaCl, 5.9 mM KCl, 11.1 mM dextrose, 25 mM NaHCO3, 1.2 mM NaH2PO4, 1.2 mM MgSO4, and 2.5 mM CaCl2). Blood gases demonstrated no significant respiratory depression, hypoxia, or acidosis (Paco2 42 ± 5 mm Hg, Pao2 > 100 mm Hg, and pH 7.40 ± 0.04). At necropsy, proper positioning of the polyethylene tubing in the CM and spread of the subarachnoid blood over the surface of the cerebral cortex were verified.
Sham-operated animals were similarly treated, and a percutaneous catheter was inserted into the CM. No blood was injected into the CM of these animals. The brain was harvested from these animals 2–3 days after the insertion of the catheter. Control animals received neither the percutaneous catheter into the CM nor subarachnoid blood injection.
Arterioles of diameters approximately 100 μm were dissected from the parietal lobes of the SAH, sham-operated, and control animals. Each vessel was placed in a vessel chamber, cannulated with dual micropipettes measuring 50–75 μm in diameter, and secured with a 10–0 suture. The vessel was continuously bathed with modified Krebs buffer, gassed with 95% O2/5% CO2 mixture, and maintained at 36.5–37.5°C and pH of 7.35–7.45. Po2 in the vessel chamber exceeded 400 mm Hg. Because the vessel was studied in a no-flow state, pressure in the micropipettes was maintained at 40 mm Hg to provide distention. The vessel was visualized, and its internal lumen diameter was measured and recorded by videomicroscopy, as previously described (9). Stability of similarly prepared vessel preparations for at least 2.5 h has been demonstrated previously (9).
Each vessel was equilibrated at 37°C for 30 min in the vessel chamber, and the baseline diameter was measured. To examine vasoconstrictive responses, the vessel was then subjected to (a) the thromboxane analog U46619 10−9–10−6 M, (b) the protein kinase C agonist 12-deoxyphorbol-13-isobutyric-20-acetate (PBE) 10−8–10−7 M, or (c) endothelin-1 10−13–10−8 M. Vessel diameter was measured at each concentration of the constrictor and %constriction calculated from the baseline diameter. Endothelial modulation of endothelin-1-mediated constriction, previously noted in coronary (10) and mesenteric (11) vessels, was verified by measuring endothelin-1 response in the presence of the NOS inhibitor NG-nitro-l-arginine (L-NNA) 10−5 M in the control vessels. At the end of each experiment, the vessel chamber was flushed with fresh Krebs buffer and the vessel re-equilibrated at 37°C. Potassium chloride (KCl) was then added to a final concentration of 100 mM, and the internal lumen diameter was measured. Functional endothelial integrity was tested by measuring the response to adenosine diphosphate (ADP) 10−5 M. Only those vessels that constricted by at least 15% to KCl and dilated significantly to ADP at the end of each experiment were considered viable and included for data analysis.
Because the response of the vessels to U46619 was found not to be influenced by SAH, we used U46619 1 μM to preconstrict additional vessels and examine their dilatory response to increasing concentrations of (a) the endothelium-dependent dilator ADP 10−9–10−4 M, (b) the endothelium-independent dilator sodium nitroprusside 10−9–10−4 M, or (c) the β-adrenergic agonist isoproterenol 10−9–10−4 M, for 2 min at each concentration. At each concentration of the dilator, the internal diameter was measured and % dilation calculated, as previously described (12). Vessel viability was tested as above.
Results from the above studies demonstrated that the cerebral microvessels had an attenuated response to the endothelium-dependent dilator ADP and an accentuated response to endothelin-1 after SAH. We then proceeded to examine if NOS3 protein and messenger RNA (mRNA) concentrations may be altered after SAH. Parietal lobe tissues from the SAH, sham-operated, and control animals were placed in liquid nitrogen (−70°C) and kept frozen until analysis. For western blotting, protein was obtained from the brain tissue by homogenizing the tissue in a lysis buffer containing 1% NP-40, 0.5% sodium deoxycholate, and 0.1% sodium dedecyl sulfate (SDS), then centrifuging at 12,000 g for 10 min at 4°C. Protein concentration of the supernatant was measured by using spectrophotometry at 595 nm (DU640, Beckman, Fullerton, CA) of an aliquot developed for 10 min in Protein Assay Dye Reagent(Bio-Rad, Hercules, CA). Total protein (30 μg/lane) was fractionated on 10% SDS-polyacrylamide gel and transferred to a polyvinylidene difluoride membrane (Immobilon-P, (Millipore, Bedford, MA). The membrane was incubated with 5% nonfat dry milk powder, 0.05% Tween20 in phosphate buffer saline (PBS) for 12 h at 4°C to block nonspecific absorption, and was then immunoblotted with the monoclonal mouse-anti-endothelial NOS3 antibody (Transduction Laboratories, Lexington, KY) 1:2000 (v/v) dilution for 2 h. After washing with PBS, the membrane was incubated for 1 h in 5% milk powder PBS containing 1:3000 diluted goat anti-mouse IgG conjugated to horseradish peroxidase (Vector Laboratories, Burlingame, CA). Peroxidase activity was visualized by using an enhanced chemiluminescence substrate system (Amersham, Arlington Heights, IL). Densitometry of digitized images of immunoprobed membranes (ScanJet 4c; Hewlett Packard) was performed by using ImageQuant software (Molecular Dynamics, Sunnyvale, CA). A monoclonal antibody raised against the structural protein Actin (Santa Cruz Biotechnology, CA) (1/1000 dilution [v/v]) was used as control (40 kDa) for equal loading. The optical density ratio of the NOS3 band to that of actin was used to compare steady levels of the protein in different groups. The level of actin does not change with SAH (13).
For quantification of NOS3 mRNA, tissues were snap frozen in liquid nitrogen and then homogenized in TriReagent (Sigma, St. Louis, MO). Twenty micrograms of total RNA were fractionated on 1% formaldehyde agarose gel and transferred on GeneScreen Plus (DuPont, Wilmington, DE) filter. The cDNA probe for NOS3 was prepared by reverse transcription-polymerase chain reaction with the sense primer corresponding to the base pairs 1010–1034 and the antisense primer corresponding to the base pairs 1469–1493 (14). The single PCR amplified product was purified by agarose gel electrophoresis, cloned into PCR™ II vector (TA Cloning Kit; Invitrogen, Carlsbad, CA). The cDNA probes were labeled with [α32 P] deoxycytidine triphosphate (New England Nuclear, Boston, MA) by using a random-priming labeling kit (Boehringer, Indianapolis, IN) and purified with the use of G-50 Quick Spin Columns (Boehringer). The typical specific probes activity used in the experiments was 1–2 × 109 cpm/μg. The blots were hybridized at 68°C for 3 h in Quickhyb solutions (Stratagene, La Jolla, CA). After the hybridization, blots were washed twice in 2× standard sodium citrate, 0.1% SDS for 15 min at 60°C. Autoradiography was performed with Kodak XAR film (Eastman Kodak, New Haven, CT) at −80°C for 48–72 h. For quantitative analysis of expression, the blots were exposed on PhosphorImager (Molecular Dynamics, Sunnyvale, CA) and analyzed by using ImageQuant software (Molecular Dynamics).
No animal contributed more than one vessel to any one experimental group. Therefore, n for each group represents the number of animals as well as the number of vessels. All data were presented as mean ± sd.
Whether there was a concentration-dependent vasomotion to a vasodilator or a constrictor was tested by using one-way analysis of variance (linear contrast). The effects of SAH on concentration response curves to the various vasomotor agents tested were analyzed by using two-way analysis of variance with a repeated measures factor, with post hoc multiple pairwise comparison (Neuman-Keuls) and stratified z tests to identify the concentrations where the differences in response were significant. Where appropriate, two-tailed Student’s t-tests were used to compare the means of two groups. P < 0.05 was considered significant. All statistics were calculated by using True Epistat software (Epistat Services, Richardson, TX).
Cerebral microvascular constrictive response to the thromboxane analog U46619 was not altered by SAH (P = 0.23) (Control:n = 7, vessel size: 102 ± 8 μm; Sham:n = 7, vessel size: 98 ± 9 μm; SAH:n = 7, vessel size: 100 ± 11 μm) (Fig. 1A), nor was the response to the protein kinase C agonist PBE (P = 0.23) (Control:n = 7, vessel size: 91 ± 8 μm; SAH:n = 7, vessel size: 92 ± 9 μm) (Fig. 1B). However, the response to endothelin-1 was insignificantly affected by sham-operation, but significantly accentuated by SAH (P < 0.001) (Control:n = 7, vessel size: 98 ± 10 μm; Sham:n = 7, vessel size: 104 ± 5 μm; SAH:n = 7, vessel size: 98 ± 17 μm) (Fig. 1C). In the control animals, the response of the cortical microvessels to endothelin-1 was increased in the presence of the NOS inhibitor L-NNA (Control with L-NNA:n = 6, vessel size: 103 ± 12 μm) (Fig. 1D).
The dilatory response of the cortical microvessels to the endothelium-dependent dilator ADP was attenuated after the sham procedure, but even more after SAH (P < 0.01 each) (Control:n = 8, vessel size: 93 ± 6 μm; Sham:n = 8, vessel size: 97 ± 5 μm; SAH:n = 8, vessel size: 95 ± 9 μm) (Fig. 2A). However, the dilatory response to the endothelium-independent dilator sodium nitroprusside was not affected by the sham procedure or SAH (P = 0.13) (Control:n = 8, vessel size: 94 ± 7 μm; Sham:n = 7, vessel size: 101 ± 5 μm; SAH:n = 8, vessel size: 90 ± 8 μm) (Fig. 2B), nor was the response to the β-adrenergic agonist isoproterenol affected by the sham procedure or SAH (P = 0.41) (Control:n = 9, vessel size: 96 ± 6 μm; Sham:n = 8, vessel size: 95 ± 8 μm; SAH:n = 8, vessel size: 104 ± 8 μm) (Fig. 2C).
NOS3 mRNA expression was increased approximately twofold in the Sham group (Control: 4.4 ± 1.1, Sham: 9.6 ± 2.4, P < 0.03) and approximately fourfold in the SAH group (SAH: 17.2 ± 0.9, P < 0.001 between control and SAH) (Fig. 3A). Despite the increased transcription of NOS3 mRNA, NOS3 protein expression was reduced in the Sham and the SAH groups compared with the Control (Fig. 3B).
The main findings of this study are 1) that in our rat model of SAH, there is an attenuation of endothelium-dependent vasodilation and accentuation of endothelin-1-mediated vasoconstriction of the microvessels after SAH, and 2) that this may be related in part to reduced protein expression of NOS3, which occurs despite an apparently compensatory increase in mRNA expression of NOS3 after SAH. Our findings suggest that endothelial dysfunction may be an important mechanism of the acute microvascular spasm after SAH and that this may be related to impaired translation of NOS3.
Our rat model of SAH is an adaptation of the SAH model of Cole et al. (7). Unlike in their model, our rats were allowed to breathe spontaneously after the subarachnoid injection of autologous blood. Nevertheless, arterial blood gases obtained 20 minutes after the injection and immediately before brain tissue harvest demonstrated that our animals had similar pH and Paco2 to those of Cole et al. (7), and that acidosis, hypoxia, and hypercarbia did not account for the observed effects in this study. A limitation of studying cerebral vasospasm in small animals is that, whereas the acute phase of vasospasm is demonstrated consistently (15), the delayed phase is not easily demonstrated (4). Using transgenic mice overexpressing CuZn-superoxide dismutase, however, Kamii et al. (16) demonstrated that significant cerebral vasospasm may be demonstrated for days after SAH in control mice compared with the transgenic animals. We chose to limit this study to the acute-phase vasospasm and, therefore, the findings of our study may be applicable to the acute-phase vasospasm only and may not be generalized to the delayed phase. Acute-phase vasospasm is predictive of mortality (17) and may be no less important than delayed-phase vasospasm. Study of the delayed phase may be more easily performed in larger animals such as primates (18) and dogs (19) and would require maintenance of the animals on mechanical ventilation for days after the induction of SAH. Finally, we have examined only microvascular dysfunction in this animal study, whereas both macrovascular and microvascular vasospasm may be important in the pathogenesis of vasospasm-induced morbidity and mortality after SAH.
Unlike the large conduit cerebral arteries in which demonstration of vasospasm by angiography has been easily reproducible (3,18), a similar demonstration in cerebral microvessels has been a challenge (5,6). In 24 patients 5–7 days after SAH, Ohkuma et al. (20) analyzed digital subtraction angiography images to distinguish between proximal cerebral circulation time (CCT), representing circulation through the large arteries, and peripheral CCT through the intraparenchymal arterioles, and correlated the findings to measure-ment of regional cerebral blood flow. They noted that in patients with no angiographically demonstrable large vessel spasm, but with a decreased regional cerebral blood flow, the peripheral CCT was prolonged, which suggests that microvascular spasm occurred after SAH.
Our study demonstrates that the cerebral microvessels have endothelial dysfunction that would predispose them to vasoconstriction and spasm. After SAH, the response of the microvessels to the endothelium-dependent dilator is attenuated. Just as in coronary (10) and mesenteric (11) vessels, endothelin-1-mediated constriction is increased in cerebral microvessels treated with the NOS inhibitor, L-NNA, which indicates that endothelin-1-mediated constriction is modulated by endothelial NO. With endothelial dysfunction such as with SAH, modulation of endothelin-1-mediated constriction is reduced, and endothelin-1 would be expected to elicit an accentuated response, just as we have demonstrated. Similar accentuation has been seen in basilar arteries (21).
Acute-stage cerebral vasospasm after SAH may be mostly related to endothelial dysfunction. Hatake et al. (22) demonstrated in human basilar arteries that within the first day after hemorrhage, i.e., during the acute vasospasmic period, endothelium-dependent dilation to various agents such as bradykinin, thrombin, and the calcium ionophore A23187 is reduced. Using a rat model of SAH, Sehba et al. (23) showed that acute vasoconstriction may be caused by the decreased availability of endothelium-derived NO, because an exogenous NO donor, N-nitroso glutathione, reversed the decrease in CBF after SAH while it had no effect in control animals. They also showed that NO metabolites are decreased in the brain regions after SAH (24). Likewise, our study adds that, in cortical arterioles, endothelium-dependent dilation and modulation of constriction are impaired shortly after SAH. Although endothelin levels may not increase in this acute stage (25), accentuated response to endothelin may contribute to pathologic changes after SAH. Whereas the endothelium is dysfunctional after SAH, the subjacent smooth muscle appears intact as it responds normally to isoproterenol or nitroprusside, as was also seen by others (26).
Our study further suggests that endothelial dysfunction in the acute stage after SAH may be related to impaired expression of NOS3 protein, even though NOS3 mRNA is increased approximately fourfold after SAH. This may be caused by either impaired translation of NOS3 or decreased stability of NOS3 protein product after SAH. Further studies will be needed to distinguish between the two. Even the sham procedure was associated with slightly decreased endothelium-dependent dilation and decreased expression of NOS3 protein despite a twofold increase in NOS3 mRNA. This may be either because the sham procedure resulted in inadvertent introduction of blood into the subarachnoid space or intracerebral NOS3 protein expression is exquisitely sensitive to any violation of the cranium and dura such as by surgery. Surgery induces endothelium-derived heme oxygenase (27) and can certainly influence enzyme expression in the endothelium.
In addition to endothelial dysfunction, various other factors may contribute to further vasomotor dysfunction and vasospasm in the delayed phase after SAH. Hemoglobin and bilirubin released after hemorrhage and clot formation can significantly reduce adenosine triphosphate levels and interfere with electron transport and creatine phosphokinase activity in vascular smooth muscle cells (28). Additionally, hemoglobin and superoxide anions are scavengers of NO (29,30); these substances may contribute to continued endothelial dysfunction. Furthermore, intracellular calcium levels increase with resultant increase in myosin light chain phosphorylation (19). Lastly, inflammatory changes follow with IgG deposition correlating with the severity and time course of vasospasm (31).
In summary, using a rat model of the acute phase of SAH, we have found that the cortical microvessels have a reduced responsiveness to the endothelium-dependent dilator ADP and an enhanced response to endothelin-1. These changes may contribute to microvascular spasm after SAH. The observed endothelial dysfunction may be related to impaired expression of NOS3 protein, which occurs despite an apparently compensatory increase in NOS3 mRNA transcription.
1. Kassell NF, Drake CG. Timing of aneurysm surgery. Neurosurgery 1982; 10: 514–9.
2. Sundt TM Jr, Whisnant JP. Subarachnoid hemorrhage from intracranial aneurysms: surgical management and natural history of disease. N Engl J Med 1978; 299: 116–22.
3. Brawley BW, Strandness DE Jr, Kelly WA. The biphasic response of cerebral vasospasm in experimental subarachnoid hemorrhage. J Neurosurg 1968; 28: 1–8.
4. Delgado TJ, Brismar J, Svendgaard NA. Subarachnoid haemorrhage in the rat: angiography and fluorescence microscopy of the major cerebral arteries. Stroke 1985; 16: 595–602.
5. Nihei H, Kassell NF, Dougherty DA, Sasaki T. Does vasospasm occur in small pial arteries and arterioles of rabbits? Stroke 1991; 22: 1419–25.
6. Katusic ZS, Milde JH, Cosentino F, Mitrovic BS. Subarachnoid hemorrhage and endothelial L-arginine pathway in small brain stem arteries in dogs. Stroke 1993; 24: 392–9.
7. Cole DJ, Nary JC, Reynolds LW, et al. Experimental subarachnoid hemorrhage in rats: effect of intravenous α-α diaspirin crosslinked hemoglobin on hypoperfusion and neuronal death. Anesthesiology 1997; 87: 1486–93.
8. Bederson JB, Pitts LH, Tsuji M, et al. Rat middle cerebral artery occlusion: evaluation of the model and development of a neurologic examination. Stroke 1986; 17: 472–6.
9. Park KW, Dai HB, Lowenstein E, Sellke FW. Vasomotor responses of rat coronary arteries to isoflurane and halothane depend on pre-exposure tone and vessel size. Anesthesiology 1995; 82: 1323–30.
10. Park KW, Lowenstein E, Dai HB, et al. Direct vasomotor effects of isoflurane in subepicardial resistance vessels from collateral-dependent and normal coronary circulation of pigs. Anesthesiology 1996; 85: 584–91.
11. Dohi Y, Luscher TF. Endothelin in hypertensive resistance arteries: intraluminal and extraluminal dysfunction. Hypertension 1991; 18: 543–9.
12. Park KW, Dai HB, Lowenstein E, et al. Isoflurane and halothane attenuate endothelium-dependent vasodilation in rat coronary microvessels. Anesth Analg 1997; 84: 278–84.
13. Oka Y, Ohta S, Todo H, et al. Protein synthesis and immunoreactivities of contraction-related proteins in smooth muscle cells of canine basilar artery after experimental subarachnoid hemorrhage. J Cereb Blood Flow Metab 1996; 16: 1335–44.
14. Gnanapandithen K, Chen Z, Kau CL, et al. Cloning and characterization of murine endothelial constitutive nitric oxide synthase. Biochim Biophys Acta 1996; 1308: 103–6.
15. Schwartz AY, Masago A, Sehba FA, Bederson JB. Experimental models of subarachnoid hemorrhage in the rat: a refinement of the endovascular filament model. J Neurosci Methods 2000; 96: 161–7.
16. Kamii H, Kato I, Kinouchi, et al. Amelioration of vasospasm after subarachnoid hemorrhage in transgenic mice overexpressing CuZn-superoxide dismutase. Stroke 1999; 30: 867–72.
17. Bederson JB, Levy AL, Ding WH, et al. Acute vasoconstriction after subarachnoid hemorrhage. Neurosurgery 1998; 42: 352–62.
18. Frazee JG. A primate model of chronic cerebral vasospasm. Stroke 1982; 13: 612–4.
19. Butler WE, Peterson JW, Zervas NT, Morgan KG. Intracellular calcium, myosin light chain phosphorylation, and contractile force in experimental cerebral vasospasm. Neurosurgery 1996; 38: 781–8.
20. Ohkuma H, Manabe H, Tanaka M, Suzuki S. Impact of cerebral microcirculatory changes on cerebral blood flow during cerebral vasospasm after aneurysmal subarachnoid hemorrhage. Stroke 2000; 31: 1621–7.
21. Zuccarello M, Romano A, Passalacqua M, Rapoport RM. Decreased endothelium-dependent relaxation in subarachnoid hemorrhage-induced vasospasm: role of ET-1. Am J Physiol 1995; 269: H1009–15.
22. Hatake K, Wakabayashi I, Kakishita E, Hishida S. Impairment of endothelium-dependent relaxation in human basilar artery after subarachnoid hemorrhage. Stroke 1992; 23: 1111–7.
23. Sehba FA, Ding WH, Chereshnev I, Bederson JB. Effects of S-nitrosoglutathione on acute vasoconstriction and glutamate release after subarachnoid hemorrhage. Stroke 1999; 30: 1955–61.
24. Sehba FA, Schwartz AY, Chereshnev I, Bederson JB. Acute decrease in cerebral nitric oxide levels after subarachnoid hemorrhage. J Cereb Blood Flow Metab 2000; 20: 604–11.
25. Gaetani P, Baena RR, Grignana G, et al. Endothelin and aneurysmal subarachnoid haemorrhage: a study of subarachnoid cisternal cerebrospinal fluid. J Neurol Neurosurg Psychiatry 1994; 57: 66–72.
26. Marshman LAG, Morice AH, Thompson JS. Increased efficacy of sodium nitroprusside in middle cerebral arteries following acute subarachnoid hemorrhage. J Neurosurg Anesthesiol 1998; 10: 171–7.
27. Motterlini R, Gonzales A, Foresti R, et al. Heme oxygenase-1-derived carbon monoxide contributes to the suppression of acute hypertensive responses in vivo. Circ Res 1998; 83: 568–77.
28. Nagatani K, Mascionpinto JE, Letarte P, et al. The effect of hemoglobin and its metabolites on energy metabolism in cultured cerebrovascular smooth-muscle cells. J Neurosurg 1995; 82: 244–9.
29. Fischer-Nakielski H, Schror K. Nitric oxide is the endothelium-derived relaxing factor in bovine pial arterioles. Stroke 1990; 21 (Suppl 12): IV46–8.
30. Shishido T, Suzuki R, Qian L, Hirakawa K. The role of superoxide anions in the pathogenesis of cerebral vasospasm. Stroke 1994; 25: 864–8.
31. Handa Y, Kabuto M, Kobayashi H, et al. The correlation between immunological reaction in the arterial wall and the time course of the development of cerebral vasospasm in a primate model. Neurosurgery 1991; 28: 542–9.