Basal Release of Nitric Oxide Is Decreased in the Coronary Circulation in Patients With Heart Failure
Abstract It is unknown whether basal release of endothelium-derived nitric oxide in the coronary artery is altered in heart failure in humans. The aim of the present study was to evaluate the effect of inhibition of nitric oxide synthesis on basal tone of the conduit and resistance coronary arteries in awake patients. Coronary blood flow velocity (Doppler guide wire) and coronary arterial diameter (quantitative coronary angiography) were measured in 14 patients with heart failure caused by nonischemic left ventricular dysfunction (7 idiopathic dilated cardiomyopathy and 7 valvular insufficiency) and 7 patients with normal ventricular function (controls). Intracoronary NG-monomethyl-l-arginine (L-NMMA), an inhibitor of nitric oxide synthesis, at graded doses decreased coronary blood flow in both groups. However, the magnitude of flow reduction was smaller in patients with heart failure than in control patients (P<.0001). The magnitude of coronary blood flow reduction in response to L-NMMA inversely correlated to indexes of left ventricular contractile function (P<.01) but was not affected by the cause of heart failure. Constriction of the large epicardial coronary artery with L-NMMA also tended to be attenuated in patients with heart failure. In summary, vasoconstricting response to L-NMMA was blunted in the coronary resistance artery in heart failure in vivo. These findings suggest that basal release of nitric oxide in the coronary circulation is decreased in patients with heart failure.
- nitric oxide
- heart failure, congestive
- coronary circulation
- cardiomyopathy, congestive
- heart valve diseases
Endothelium-derived nitric oxide (NO) is an endogenous potent vasodilator that is released from vascular endothelial cells in response to various vasoactive substances as well as mechanical shear stress exerted on the endothelial cell.1 2 3 Since its discovery, the role of NO in regulating blood flow under physiological and pathological conditions such as hypercholesterolemia4 and hypertension5 has been vigorously investigated.
As for congestive heart failure (CHF), Habib et al6 reported that systemic administration of the NO synthase inhibitor NG-monomethyl-l-arginine (L-NMMA) caused systemic vasoconstriction, as evidenced by increased vascular resistance, in patients with heart failure and that this response was greatest in those with highest basal resistance. In the forearm circulation, the constricting response to intra-arterial L-NMMA in heart failure patients was reportedly similar or even enhanced compared with control patients.7 8 This evidence would suggest that basal release of NO is at least preserved and plays a substantial role in the peripheral circulation with heart failure. However, NO synthesis and its contribution to basal arterial tone may be heterogeneously affected in different vascular territories in heart failure.9 10
Whether basal or agonist-mediated release of NO in the coronary circulation is affected in CHF remains to be determined. Data obtained in animal experiments are inconsistent regarding the functional role of NO in regulating coronary blood flow in heart failure. O’Murchu and coworkers9 demonstrated that coronary artery rings from dogs with pacing-induced heart failure showed enhanced dilator responses to α2-adrenergic agonists that were mediated by NO. In contrast, Wang et al11 observed depressed NO-mediated controls of the coronary circulation in dogs with heart failure as evidenced by attenuated dilation of the large epicardial coronary artery in response to brief coronary occlusion and acetylcholine. However, no study has evaluated the physiological role of NO in the regulation of coronary blood flow in patients with heart failure under either baseline or stressed conditions. Thus, in the present study we aimed to elucidate the role of basally released NO in the control of coronary blood flow in patients with CHF caused by noncoronary artery disease. To achieve this, we examined the effects of intracoronary administration of a specific inhibitor of NO synthesis, L-NMMA, on coronary blood flow and diameter in patients with heart failure and control patients.
Fourteen patients with heart failure caused by nonischemic left ventricular dysfunction (mean age, 57±11 years; 8 men and 6 women; CHF group) were studied (Table 1⇓). The cause of left ventricular dysfunction was idiopathic dilated cardiomyopathy in 7 patients (DCM group; patients 1 through 7 in Table 1⇓) and volume-overload left ventricular dysfunction caused by valvular insufficiency in 7 patients (VHD group; patients 8 through 14). All patients had been admitted for the treatment of symptomatic heart failure; New York Heart Association functional class on admission was II in 2 patients, III in 5, and IV in 7. The study protocol was performed 1 to 2 weeks after initial management of heart failure. All patients had normal coronary arteriograms. We studied 7 other patients without clinical evidence of left ventricular dysfunction (mean age, 58±8 years; 4 men and 3 women) who had been referred to our hospital for evaluation of chest pain (control group, Table 1⇓). All control patients had negative treadmill testing and normal coronary angiograms. Clinical characteristics of the studied patients and baseline hemodynamic variables and indexes of left ventricular function in the studied groups are shown in Tables 1⇓ and 2⇓, respectively. Left ventricular ejection fraction of DCM patients was 33±12% (ranging 16% to 46%, P<.0001 versus controls), whereas that of VHD patients was 63±7%. It has been well recognized that ejection phase indexes of left ventricular contraction such as ejection fraction are often within the normal range in patients with contractile dysfunction due to valvular insufficiency.12 13 To obtain an index of left ventricular systolic function that is more independent of load and can be clinically estimated with relative ease, we additionally determined the quotient of end-ejection arterial pressure and left ventricular end-systolic volume index (Pee/LVESVI). This variable is an approximation of left ventricular end-systolic elasticity,14 15 which has been experimentally and clinically shown to be relatively independent of load within a broad range of preload and afterload.16 As shown in Table 2⇓, the Pee/LVESVI ratio was significantly smaller in both CHF subgroups (DCM and VHD groups) compared with that in the control group. Cardiac index was significantly (P<.05) reduced in VHD patients. One DCM patient (patient 1) had diabetes with a positive glucose tolerance test, and 1 patient with valvular insufficiency (patient 9) was hyperlipidemic (total cholesterol >260 mg/dL or 6.72 mmol/L). Two control group patients (patients 1 and 5) had one or two risk factors. In summary, no coronary risk factor was found in 6 of 7 DCM patients, 6 of 7 VHD patients, and 5 of 7 control patients.
The study protocol was approved by the Institutional Review Committee on Human Research, Faculty of Medicine, Kyushu University. Written informed consent was obtained from each patient before the study.
Cardiac catheterization was performed with patients in the fasting state after 5 mg oral diazepam. All cardiovascular medications were discontinued at least 24 hours before the study. Right and left heart catheterization was performed via the femoral approach. Biplane left ventriculograms at the right and left oblique projections were recorded with a 6F pigtail catheter and power injector. At least 20 minutes after the left ventriculography, the following protocols were performed.
First, we examined the effects of the NO synthesis inhibitor L-NMMA (Clinalfa) on the following variables. L-NMMA at doses of 50, 50, and 100 μmol (cumulative doses of 50, 100, and 200 μmol) was administered sequentially. The estimated maximum molar concentration of L-NMMA in our dosing protocol would be 590 μmol/L if resting coronary blood flow of the left coronary artery is assumed to be 170 mL/min. This dose of L-NMMA has been shown to be effective in inhibiting NO synthesis activity in isolated coronary arteries.17 Furthermore, we have demonstrated that this dosing protocol of L-NMMA inhibits acetylcholine (10 μg/min)–induced dilation of large epicardial coronary arteries and increases in coronary blood flow in patients with normal coronary angiograms.18 Systemic arterial pressure, heart rate, 12-lead electrocardiogram, and coronary blood flow velocity at the proximal left anterior descending artery (intracoronary Doppler guide wire) were continuously measured throughout the study. Each dose of L-NMMA was diluted in 5% glucose solution and infused manually over 1 minute into the left coronary artery via a Judkins-type left coronary catheter. At least 2 minutes was allowed to elapse before administration of the next dose of L-NMMA. Coronary angiography was performed at baseline (ie, before L-NMMA administration) and after the final dose of L-NMMA was given to obtain the diameter of the large epicardial coronary artery. Second, papaverine (10 mg) was administered into the left coronary artery to obtain coronary flow reserve under maximal coronary vasodilation.19 Coronary flow reserve was defined as the ratio of maximal coronary blood flow after papaverine to baseline flow.
Quantitative Coronary Angiography and Measurement of Coronary Blood Flow Velocity
Quantitative coronary angiography was performed with a Siemens cineangiographic system (Bicor & Hicor), as previously described.20 21 Nonionic contrast material (iomeprol, Eisai) was used. An appropriate view was selected that allows clear visualization of the vessel under study. Throughout the study, the angle of projection, the distance from the x-ray focus to the object, and that from the object to the image intensifier were kept constant. An end-diastolic frame of the coronary angiogram was selected, and the luminal diameter of the segment of the artery distal to the Doppler guide wire tip was determined. The tip of the Judkins catheter with known diameter was used as a reference to obtain the absolute diameter of the vessel. The accuracy and precision of our quantitative angiographic system were validated with precision-drilled models, as previously reported.20 Measurements were made three times, and the averaged value was used for analysis. Interobserver and intraobserver reproducibilities were high (r=.96 and r=.98, respectively).
Coronary blood flow velocity was measured with a 0.014-inch Doppler guide wire (FloWire, Cardiometrics Inc) and a fast Fourier transform–based spectral analyzer (FloMap). A Doppler guide wire was advanced via the Judkins catheter, and the tip of the Doppler guide wire was positioned at the proximal portion of the left anterior descending coronary artery. Peak coronary blood flow velocity signal was continuously determined with a spectral analyzer and recorded on a multichannel recorder (Nihon-Koden). Steady-state signals were obtained, and the value based on at least three consecutive beats was used for analysis. Volumetric coronary blood flow was calculated at baseline and after administration of the final dose of L-NMMA using the following formula22 : Coronary Blood Flow (mL/min)=0.5×Averaged Peak Velocity (cm/min)×Cross-sectional Area (cm2).
Coronary vascular resistance index was arbitrarily defined as the quotient of mean arterial pressure (millimeters of mercury) and coronary blood flow velocity (centimeter per second) and is presented as a percentage of the baseline value. Changes in diameter and coronary blood flow in response to L-NMMA are expressed as the percent change from the baseline value.
Left ventricular volume and ejection fraction were determined by the area-length analysis of biplane cine ventriculograms. End-systolic volume index was calculated as left ventricular volume at end systole divided by the patient’s body surface area.
Data are expressed as mean±SD, unless otherwise indicated. Comparisons of baseline systemic and coronary hemodynamic variables within groups were performed by ANOVA followed by Bonferroni’s multiple comparison test if indicated. Effects of L-NMMA on systemic and coronary hemodynamics within each group of patients were analyzed by Student’s paired t test. Changes caused by L-NMMA in coronary diameter and blood flow between the groups were compared by unpaired t test. A simple linear regression analysis was used to analyze relations between changes in coronary blood flow and indexes of left ventricular function. Differences were considered statistically significant at a value of P<.05.
Effect of L-NMMA on Systemic Hemodynamics
Baseline heart rate, mean aortic pressure, and pressure-rate product (Systolic Arterial Pressure×Heart Rate) did not differ significantly among the three groups (Table 3⇓). L-NMMA caused a slight but significant (P<.05) decrease in heart rate in the control group (70±7 to 66±7 beats per minute) but not in the CHF subgroups (Table 3⇓). L-NMMA increased mean arterial pressure slightly but significantly in the control group (92±13 to 96±12 mm Hg, P<.05) but not in the CHF subgroups. Pressure-rate product was unchanged after L-NMMA in the DCM, VHD, and control groups.
Effect of L-NMMA on Epicardial Coronary Diameter
L-NMMA decreased the luminal diameter of the left descending coronary artery by 6.4%, from 3.14±0.89 to 2.94±0.88 mm (P<.05) in control patients (Table 3⇑, Fig 1A⇓). L-NMMA did not significantly change luminal diameter in either CHF subgroup.
Effect of L-NMMA on Coronary Vascular Resistance Index and Blood Flow
L-NMMA progressively increased coronary vascular resistance index in a dose-dependent manner in the control group (P<.0001) as well as the CHF group (P<.05) (Fig 1B⇑). However, constricting responses to L-NMMA were significantly attenuated in the DCM and VHD groups (both P<.01 versus control).
L-NMMA decreased coronary blood flow from 53±21 to 35±15 mL/min (P<.001) in control patients. L-NMMA also decreased coronary blood flow from 56±39 to 50±35 mL/min (P<.05) in DCM patients and from 57±26 to 51±24 mL/min (P<.01) in VHD patients. Percent changes in coronary blood flow induced by L-NMMA were significantly (P<.0001) smaller in CHF than control patients (DCM, –8±7%; VHD, –11±6%; control, –34±4%; Fig 1C⇑). There was a statistically significant correlation between the L-NMMA–induced changes in coronary blood flow and left ventricular ejection fraction (P<.01, r=–.620) as well as between the L-NMMA–induced changes in coronary blood flow and the Pee/LVESVI ratio (P<.001, r=–.751) (Fig 2⇓).
Effect of Papaverine
Coronary flow reserve as assessed by intracoronary papaverine was not significantly different between the CHF and control groups (2.8±0.9 versus 3.5±0.4, respectively). There was no significant relation between the L-NMMA–induced decrease in coronary blood flow and coronary flow reserve.
The basal release of NO in the coronary circulation or its physiological role in regulating coronary blood flow has not been determined in awake patients with heart failure. The novel finding in the present study is that the coronary constricting effect of intracoronary L-NMMA was less in patients with heart failure but was not related to the cause of heart failure. These results support the hypothesis that basal NO production, release, or both are decreased in the coronary circulation of the human failing heart in vivo.
Recent clinical investigations suggest that basal release of NO is at least preserved and plays a substantial role in the peripheral circulation in patients with heart failure.6 7 8 However, NO synthesis and its contribution to basal arterial tone may be heterogeneously affected in different vascular territories in heart failure.9 10 In the coronary circulation, however, animal studies have yielded conflicting results. In rats with large infarction, Drexler et al10 demonstrated that the vasoconstrictor response to L-NMMA was blunted, suggesting decreased basal release of NO. In the canine model with heart failure induced by chronic tachypacing, it also has been reported that the coronary dilator response to acetylcholine and vasodilation elicited by cardiac chemoreflex and carotid receptors, all of which are known to be NO dependent, were impaired after the development of heart failure.11 23 In contrast, O’Murchu et al9 reported that in large epicardial coronary artery excised from dogs with pacing-induced heart failure, NO production in response to α2-adrenergic stimulation was enhanced. Furthermore, the possibility has been suggested that increased cytokines, such as tumor necrosis factor-α, in heart failure upregulate an inducible form of NO synthase in the heart and increase NO production.24 Therefore, we designed the present study to investigate the role of NO in the regulation of baseline coronary blood flow in patients with heart failure.
Basal Release of NO in the Coronary Circulation With Heart Failure
NO is continuously produced and released by the endothelium and plays an important role in the regulation of vasomotion and therefore organ perfusion. We and others have reported that intracoronary infusion of an l-arginine analogue such as L-NMMA causes vasoconstriction of large epicardial and small resistance arteries in humans.18 25 Our finding that L-NMMA reduced the luminal diameter of the large epicardial coronary artery and decreased coronary blood flow in control patients is consistent with those previous studies and suggests a tonic dilator effect of basally released NO on the conduit and resistance coronary arteries.
We demonstrated that the reduction in coronary blood flow in response to L-NMMA was significantly less in patients with heart failure. Since L-NMMA did not change myocardial oxygen consumption, as evidenced by an unaltered pressure-rate product in both groups, our observation indicates that the constricting response of the resistance coronary arteries to L-NMMA is blunted in patients with heart failure. Coronary risk factors such as hypercholesterolemia, diabetes, smoking, and hypertension are known to be associated with endothelial dysfunction1 2 4 26 and reduced bioavailability of NO25 irrespective of heart failure. However, in the present study, only one DCM patient had concomitant diabetes and a current smoking habit, and one VHD patient was hyperlipidemic (total cholesterol >260 mg/dL or 6.72 mmol). Thus, it is unlikely that the difference in clinical characteristics other than heart failure alone determined our results. Intriguingly, the blunted response to L-NMMA was most prominent in patients with more advanced left ventricular dysfunction. This may suggest that basal release of NO from the coronary circulation is relevant to the progression of heart failure.
Although we did not examine the mechanisms by which the constricting response to L-NMMA is attenuated in heart failure, there may be several possibilities. First, the production of NO from the coronary vessels may be decreased. Recent findings that the production of nitrite was decreased in excised coronary arteries from the canine and human failing heart11 27 support this hypothesis. Shear stress exerted on the endothelium is known to facilitate NO release.2 26 28 Pulse frequency (ie, heart rate), pulse pressure, and coronary blood flow velocity were not different in the studied groups. We do not, however, exclude the possibility that reduced shear stress in heart failure might have attenuated basal release of NO. Second, it should be considered whether the attenuated response to L-NMMA observed in the present study was due to the difference in baseline coronary tone. If the basal coronary tone had been already increased in our CHF patients, further constrictor response to L-NMMA could have been smaller than in control patients. We did not examine responses to other vasoconstrictor agents in these patients. However, it has been suggested that vasoconstricting responses are generally augmented but not attenuated in heart failure.29 30 In the present study, we observed comparable vasodilator responses to intracoronary papaverine and baseline vascular resistance, which may suggest that basal coronary tone was comparable in our patients with and without heart failure. Thus, it is unlikely that the attenuated response to L-NMMA was accounted for by the difference in baseline coronary tone. It also has been reported that ouabain, a cardiac glycoside, attenuates acetylcholine-induced endothelium-dependent vasodilation in human subcutaneous arteries in vitro.31 We do not know whether digoxin may account for the attenuated response to L-NMMA in our patients. Digoxin had been taken by nine of the CHF patients but none of the control patients. However, other studies have demonstrated that ouabain does not inhibit acetylcholine-induced relaxation in the coronary artery.32 The latter observation suggests that the effect may be heterogeneous in different vascular beds. Indeed, the response to L-NMMA in our CHF patients with and without digoxin treatment was not significantly different. In any case, our present observation supports the hypothesis that basal release of NO may not be increased in the coronary circulation with heart failure.
Effect of L-NMMA on the Diameter of the Large Epicardial Coronary Artery
The reduction in the luminal diameter of the large epicardial coronary artery in response to intracoronary L-NMMA tended to be greater in the control patients. This decrease in the coronary diameter must have been at least partly secondary to the decrease in coronary blood flow by L-NMMA. It is therefore unknown whether the direct vasoconstricting effect of L-NMMA on the large epicardial coronary artery is also blunted in heart failure.
Study limitations include the following. First, we used 200 μmol L-NMMA cumulatively in the present study. This caused a slight but significant rise in arterial pressure, and higher dose of the drug would have caused systemic hypertension. To avoid secondary changes in coronary blood flow caused by elevated arterial pressure and hence increased myocardial oxygen demand, we did not test a higher dose. However, the constrictor effect of L-NMMA plateaued at doses higher than 50 μmol in CHF patients (Fig 1B⇑); therefore, it is unlikely that a higher dose of L-NMMA might have caused larger responses. Second, we did not evaluate the stimulated release of NO in our patients. Studies from other laboratories have shown that acetylcholine-induced coronary vasodilation is impaired in dilated cardiomyopathy.33 34 Responses to other endothelium-dependent stimuli such as substance P remain to be elucidated. Third, it is not known from our study how the attenuated basal (and stimulated) release of NO may influence the control of coronary blood flow under physiologically stressed conditions such as exercise or tachypacing. Future studies are warranted to address these questions.
Intracoronary L-NMMA decreased baseline coronary blood flow in patients with and without heart failure in vivo. The vasoconstricting effect of L-NMMA was less in CHF patients. Our data support the hypothesis that basal release of NO in the coronary circulation is decreased in patients with heart failure caused by noncoronary artery disease.
- Received September 9, 1996.
- Revision received October 11, 1996.
- Accepted December 31, 1996.
Kiowski W. Endothelial function in humans: studies of forearm resistance vessels. Hypertension. 1991;18(suppl II):II-84-II-89.
Egashira K, Hirooka Y, Kai H, Sugimachi M, Suzuki S, Inou T, Takeshita A. Reduction of serum cholesterol with pravastatin improves endothelium-dependent coronary vasomotion in patients with hypercholesterolemia. Circulation. 1994;89:2519-2524.
Koller A, Huang A. Impaired nitric oxide-mediated flow-induced dilation in arterioles of spontaneously hypertensive rats. Circ Res. 1994;74:416-421.
O’Murchu B, Miller VM, Perrella MA, Burnett JCJ. Increased production of nitric oxide in coronary arteries during congestive heart failure. J Clin Invest. 1994;93:165-171.
Drexler H, Hablawetz E, Wenyan L, Riede U, Christes A. Effects of inhibition of nitric oxide formation on regional blood flow in experimental myocardial infarction. Circulation. 1992;86:255-262.
Wang J, Seyedi N, Xu X-B, Wolin MS, Hintze TH. Defective endothelium-mediated control of coronary circulation in conscious dogs after heart failure. Am J Physiol. 1994;266:H670-H680.
Borow KM, Green LH, Mann T, Sloss LJ, Braunwald E, Collins JJ Jr, Cohn L, Grossman W. End-systolic volume as a predictor of postoperative left ventricular performance in volume overload from valvular regurgitation. Am J Cardiol. 1980;68:655-663.
Harasawa Y, Yoshida T, Sunagawa K, Takeshita A. A novel index of left ventricular systolic function that is less afterload-dependent than ejection fraction. Heart Vessels. 1992;suppl 8:15. Abstract.
McKay RG, Aroesty JM, Heller GV, Royal HD, Warren SE, Grossman W. Assessment of the end-systolic pressure-volume relationship in human beings with the use of a time-varying elastance model. Circulation. 1986;74:97-104.
Sagawa K. The ventricular pressure-volume diagram revisited. Circ Res. 1978;43:677-687.
Tschudi M, Richard V, Bühler FR, Lüscher TF. Importance of endothelium-derived nitric oxide in porcine coronary resistance arteries. Am J Physiol. 1991;260:H13-H20.
Egashira K, Katsuda Y, Mohri M, Kuga T, Tagawa T, Kubota T, Hirakawa Y, Takeshita A. Role of endothelium-derived nitric oxide in coronary vasodilation induced by pacing tachycardia in humans. Circ Res. 1996;79:331-335.
Kuga T, Egashira K, Mohri M, Tsutsui H, Harasawa Y, Urabe Y, Ando S, Shimokawa H, Takeshita A. Bradykinin-induced vasodilation is impaired at the atherosclerotic site but is preserved at the spastic site of human coronary arteries in vivo. Circulation. 1995;92:183-189.
Doucette JW, Corl PD, Payne MH, Flynn AE, Goto M, Nassi M, Segal J. Validation of a Doppler guide wire for intravascular measurement of coronary artery flow velocity. Circulation. 1992;85:1899-1911.
Zhao G, Shen W, Xu X, Ochoa M, Bernstein R, Hintze TH. Selective impairment of vagally mediated, nitric oxide-dependent coronary vasodilation in conscious dogs after pacing-induced heart failure. Circulation. 1995;91:2655-2663.
Haywood GA, Tsao PS, von der Leyen HE, Mann MJ, Keeling PJ, Trindade PT, Lewis NP, Byrne CD, Rickenbacher PR, Bishopric NH, Cooke JP, McKenna WJ, Fowler MB. Expression of inducible nitric oxide synthase in human heart failure. Circulation. 1996;93:1087-1094.
Quyyumi AA, Dakak N, Andrews NP, Husain S, Arora S, Gilligan DM, Cannon RO. Nitric oxide activity in the human coronary circulation: impact of risk factors for coronary atherosclerosis. J Clin Invest. 1995;95:1747-1755.
Antony I, Lerebours G, Nitenberg A. Loss of flow-dependent coronary artery dilatation in patients with hypertension. Circulation. 1995;91:1624-1628.
Kichuk MR, Seyedi N, Zhang X, Marboe CC, Michler RE, Addonizio LJ, Kaley G, Nasjletti A, Hintze TH. Regulation of nitric oxide production in human coronary microvessels and the contribution of local kinin formation. Circulation. 1996;94:44-51.
Canty JM Jr, Schwartz JS. Nitric oxide mediates flow-dependent epicardial coronary vasodilation to changes in pulse frequency but not mean flow in conscious dogs. Circulation. 1994;89:375-384.
Teerlink JR, Gray GA, Clozel M, Clozel JP. Increased vascular responsiveness to norepinephrine in rats with heart failure is endothelium dependent: dissociation of basal and stimulated nitric oxide release. Circulation. 1994;89:393-401.
Woolfson RG, Poston L. Effect of ouabain on endothelium-dependent relaxation of human resistance arteries. Hypertension. 1991;17:619-625.
Treasure CB, Vita JA, Cox DA, Fish RD, Gordon JB, Mudge GH, Colucci WS, St John Sutton MG, Selwyn AP, Alexander RW, Ganz P. Endothelium-dependent dilation of the coronary microvasculature is impaired in dilated cardiomyopathy. Circulation. 1990;81:772-779.