to maintain systemic blood pressure5 and have beendemonstrated in critically ill patients with myocardialinfarction,6 cardiogenic shock,7 septic shock,8 adult respi-ratory distress syndrome,9 heart failure,10,11 and acute kid-ney failure.12 To date, several animal and human studies ofseptic shock have suggested that endothelin release may bepathologic13-18 or homeostatic.18,19 Increased plasmaendothelin levels have been demonstrated in patientsundergoing nonruptured abdominal aortic aneurysm(AAA) repair with infrarenal20,21 and supraceliac aorticcross-clamping,22 as well as in one animal model ofinfrarenal aortic clamping and subsequent exsanguina-tion.23 To date, however, there are no reports of theendothelin response in patients undergoing repair of rup-tured AAA. The aim of this study was to examine, for thefirst time, perioperative changes in plasma levels of big ET-1 and ET-1 in patients undergoing repair of rupturedAAA. We hypothesized that hemorrhagic shock, ischemia,and reperfusion would lead to increased synthesis andsecretion of endothelin, which would predispose to thedevelopment of organ failure, one of the principal causesof death in this group of patients.

METHODS

Patients. Fourteen consecutive patients (13 men and1 woman of median age 74 years; range, 65-86) whounderwent repair of ruptured infrarenal AAA and survivedto at least 24 hours after surgery were prospectively stud-ied. Lothian Research Ethics Committee approval was

Endothelin-1 (ET-1) is the most potent known vaso-constrictor. It is principally secreted abluminally from vas-cular endothelial cells but may enter the circulation ifconcentrations are high at the endothelial cells–vascularsmooth muscle interface.1 ET-1 leads to vasoconstrictionin resistance vessels, especially the coronary, cerebral, andrenal circulation, by acting on ETA receptors in vascularsmooth muscle cells, and ETB receptors on vascularsmooth muscle and endothelial cells. Big ET-1 is theimmediate precursor of ET-1, and its conversion to bio-logically active ET-1 by endothelin-converting enzymesoccurs mainly in the vessel wall.2,3 Big ET-1 is detectablein the plasma for considerably longer than ET-1,4 andincreased plasma levels of big ET-1 are considered to rep-resent increased ET-1 generation. Elevated plasmaendothelin levels may form part of a homeostatic response

1242

From the Vascular Surgery Unit, Royal Infirmary of Edinburgh,a ClinicalPharmacology Unit, University Department of Medical Sciences,Western General Hospital,b and University Department of VascularSurgery, Heartlands Hospital.c

Competition of interest: nil.Supported by a joint Royal College of Surgeons of Edinburgh and British

Union Provident Association fellowship.Reprint requests: Andrew W. Bradbury, Bsc, MD, University Department

of Vascular Surgery, Heartlands Hospital, Bordesley Green East,Birmingham B9 5SS, United Kingdom.

Published online Apr 12, 2001.Copyright © 2001 by The Society for Vascular Surgery and The American

Association for Vascular Surgery.0741-5214/2001/$35.00 + 0 24/1/113296doi:10.1067/mva.2001.113296

Plasma endothelin levels and outcome in patientsundergoing repair of ruptured infrarenalabdominal aortic aneurysmDonald J. Adam, FRCSEd,a Stephen M. Evans, BSc, FRCSEd,a David J. Webb, MD, FRCP,b andAndrew W. Bradbury, BSc, MD, FRCSEd,c Edinburgh and Birmingham, United Kingdom

Background: Endothelin-1 (ET-1) is the most potent known vasoconstrictor. Elevated plasma levels have been demon-strated in patients with myocardial infarction, cardiogenic and septic shock, and respiratory, heart, and kidney failure,as well as in those undergoing elective abdominal aortic aneurysm (AAA) repair. However, endothelin levels have notpreviously been examined in patients undergoing repair of ruptured AAA. We hypothesized that hemorrhagic shock,lower torso ischemia, and reperfusion associated with ruptured AAA repair lead to increased synthesis and secretion ofET-1, which, in turn, predispose to organ failure, one of the principal causes of death in this condition.Methods: Fourteen patients were studied. Plasma levels of big ET-1 and ET-1 were measured immediately before oper-ation and immediately before, 5 minutes, and 6 hours after aortic clamp release. Results: All patients survived for at least 24 hours after operation. Big ET-1 levels were above the normal range atone or more sample points in all patients, and the ET-1 levels were above the normal range in all survivors and fourof five nonsurvivors. Five patients who died of organ failure had significantly lower big ET-1 levels at all samplepoints and significantly lower ET-1 levels after 5 minutes of reperfusion when compared with survivors. PreoperativeET-1 levels were significantly lower in eight patients who subsequently developed kidney failure than in six patientswho did not.Conclusion: Contrary to our original hypothesis, these novel data demonstrate that patients with ruptured AAA inwhom fatal postoperative organ failure develops have significantly lower perioperative endothelin levels than survivors.(J Vasc Surg 2001;33:1242-6.)

obtained. The median (range) delay between the onset ofsymptoms of rupture and hospital admission was 5 (3-14)hours. All patients had at least one documented episode ofhypotension (systolic blood pressure < 100 mm Hg)before surgery.

Operative methods. Ruptured AAA was defined bythe presence of fresh retroperitoneal blood at operation.No patient had intraperitoneal rupture. All patients under-went AAA repair under general anesthesia through a trans-verse supraumbilical incision with infrarenal aorticcross-clamping. No patient required suprarenal aorticclamping. An aortoaortic graft was inserted in 10 patients,an aortobifemoral graft in three, and an aortobiiliac graftin one patient.

Sample collection and assay methods. Blood (8.2mL) was obtained from an indwelling radial arterial lineand placed into ethylenediamine tetra-acetic acid (2.7 mL)for estimation of hematocrit, and lithium heparin (5.5mL) for estimation of big ET-1 and ET-1. Blood was sam-pled immediately before the induction of anesthesia (sam-ple A), immediately before aortic clamp release (sampleB), and 5 minutes (sample C) and 6 hours (sample D)after aortic clamp release. Samples were placed immedi-ately on ice and spun in a centrifuge within 30 minutes ofcollection at 1400g for 30 minutes at 4°C. Plasma wasseparated and stored at –70°C for later batch analysis.

Hematocrit was estimated in the routine hematologylaboratory with the fully automated Sysmex NE 8000 ana-lyzer (Sysmex, Milton Keynes, UK). Plasma immunoreac-

JOURNAL OF VASCULAR SURGERYVolume 33, Number 6 Adam et al 1243

tive big ET-1 and ET-1 concentrations were measured byuse of an acetic acid extraction technique24 and a modifiedcommercial radioimmunoassay with rabbit antihuman bigET-1 or ET-1 (Peninsula Laboratories Europe, St Helens,UK). Sample extract was incubated with either big ET-1or ET-1 antibody for 24 hours at 4°C. After incubation,125-labeled big ET-1 (Peninsula Laboratories Europe) orET-1 (NEN Life Science Products, Boston, Mass) wasadded, and incubation was continued for an additional 20minutes at 4°C. Complexes were precipitated withAmerlex donkey antirabbit antibody (Amersham LifeSciences Limited, UK) and counted for radioactivity. Thelower limits of detection for the big ET-1 and ET-1 assaysare 1 pg/mL and 0.25 pg/mL, respectively. The normallaboratory range for big ET-1 is 10 to 60 pg/mL, and forET-1 1.5 to 4.5 pg/mL.

Definitions of postoperative organ failure. Cardiacfailure was defined as arrhythmia requiring pharmacologictreatment to maintain cardiovascular stability and/or sus-tained periods of hypotension (mean arterial pressure ≤ 60mm Hg) requiring fluid resuscitation and inotropic sup-port. Respiratory failure was defined as hypoxia requiringmechanical ventilatory support for more than 4 days.Kidney failure was defined as elevated serum creatininelevel greater than or equal to 250 µmol/L and/or therequirement for renal replacement therapy. Disseminatedintravascular coagulation was defined as clinical evidenceof hemorrhage accompanied by laboratory evidence ofthrombocytopenia, prolonged clotting times, hypofibrino-

Fig 1. Individual data points for big ET-1 immediately beforeinduction of anesthesia (sample A), immediately before release ofaortic clamp, (sample B), and 5 minutes (sample C), and 6 hours(sample D) after aortic clamp release in 14 patients who under-went operation for ruptured AAA. Survivors (n = 9) are repre-sented by black squares and nonsurvivors (n = 5) by black circles.Normal laboratory range for big ET-1 (10-60 pg/mL) is shownby parallel horizontal lines. Mann-Whitney U test was used to testwhether medians of samples in survivors and nonsurvivors weresignificantly different from each other at each sample point. A Pvalue less than .05 was regarded as statistically significant.

Fig 2. Individual data points for ET-1 immediately before induc-tion of anesthesia (sample A), immediately before release of aor-tic clamp, (sample B), and 5 minutes (sample C), and 6 hours(sample D) after aortic clamp release in 14 patients who under-went operation for ruptured AAA. Survivors (n = 9) are repre-sented by black squares and nonsurvivors (n = 5) by black circles.Normal laboratory range for ET-1 (1.5-4.5 pg/mL) is shown byparallel horizontal lines. Mann-Whitney U test was used to testwhether medians of samples in survivors and nonsurvivors weresignificantly different from each other at each sample point. A Pvalue less than .05 was regarded as statistically significant. NS,Not significant.

genemia, and elevated levels of fibrin/fibrinogen degrada-tion products.

Statistical methods. The Mann-Whitney U test wasused to test whether the medians of samples in survivorsand nonsurvivors were significantly different from eachother and to examine whether there was a difference inlevels between patients who had postoperative organ fail-ure compared with patients who did not. The Kruskal-Wallis one-way analysis of variance was used to examinewhether assay levels changed significantly over the foursampling points in survivors and nonsurvivors. A P valueof less than .05 was regarded as statistically significant.

RESULTS

Clinical data. All patients were admitted to theintensive therapy unit after operation. All patients survivedfor at least 24 hours after operation, but five (36%) died inthe postoperative period. Clinical and operative data forsurvivors and nonsurvivors are summarized in Table I, andpostoperative complications are shown in Table II.

Plasma levels of big ET-1 and ET-1. There was nosignificant difference in hematocrit between survivors andnonsurvivors at any of the sample points. The values of bigET-1 and ET-1 in survivors and nonsurvivors are shown inFigs 1 and 2, respectively. The big ET-1 level was abovethe normal laboratory range at one or more samplingpoints in all patients, and the ET-1 level was above thenormal range in all survivors and four of five nonsurvivors.When compared with nonsurvivors, survivors had signifi-cantly higher levels of big ET-1 at all four sampling pointsand significantly higher levels of ET-1 after 5 minutes’reperfusion than nonsurvivors. When compared with pre-operative levels, there was a significant increase in big ET-1 levels after 6 hours of reperfusion in survivors. Innonsurvivors, there was a significant increase in ET-1 lev-els between 5 minutes and 6 hours after reperfusion.Preoperative ET-1 levels were significantly lower in eight

patients who subsequently had kidney failure (median,3.72; range, 2.76-6.0 pg/mL) than in six patients who didnot (median, 5.89; range, 3.86-7.23 pg/mL; P = .02).There was no significant difference in big ET-1 or ET-1levels between patients who did and did not have cardiacfailure, respiratory failure, or disseminated intravascularcoagulation.

DISCUSSION

This study is the first to examine the relationshipbetween perioperative endothelin levels, organ failure, anddeath in patients undergoing ruptured AAA repair. Theprincipal finding is that, contrary to our original hypothe-sis, patients who died had significantly lower perioperativeendothelin levels than survivors.

Previous studies of the endothelin response to lowertorso ischemia and reperfusion are few and contradictory.Antonucci et al20 examined the effect of intraoperativenifedipine infusion on endothelin-dependent renal vaso-constriction in five patients undergoing nonrupturedinfrarenal AAA repair and demonstrated a transient butsignificant increase in plasma ET-1 and -2 levels at the endof the period of aortic cross-clamping. The authors con-cluded that nifedipine prevented the renal vasoconstrictorresponse to endothelin because there was no significantdifference in creatinine clearance and glomerular filtrationrate after operation compared with before operation.Fukuda et al21 measured arterial and iliac vein ET-1 levelsin seven patients undergoing elective aortic aneurysmrepair. There was no significant change in arterial ET-1levels, but a significant increase in iliac vein ET-1 levelsoccurred immediately after aortic clamp release and perfu-sion of the first limb. Venous ET-1 levels showed a signif-icant correlation with venous O2 content, pH, partialpressure of oxygen O2, O2 saturation, and base excess sug-gesting that ET-1 production occurred because of lowerlimb ischemia. Lintott et al22 were the first to attempt to

JOURNAL OF VASCULAR SURGERY1244 Adam et al June 2001

Table I. Clinical and operative data

Survivors (n = 9) Nonsurvivors (n = 5)

Median Range Median Range P value*

PreoperativeDuration of symptoms (h) 4 (3-14) 6 (5-12) NSSerum creatinine level (µmol/L) 138 (82-176) 115 (77-183) NSIntravenous fluid administration before clamping (L) 0.5 (0.2-5.5) 0.7 (0.1-1.0) NS

IntraoperativeTotal operation time (min) 140 (75-240) 105 (75-200) NSAortic clamp time (min) 90 (40-185) 75 (55-135) NSMeasured blood loss (L) 4.0 (1.0-6.4) 3.3 (1.5-11.0) NSCrystalloid and colloid administration (L) 3.4 (1.5-8.0) 3.8 (3.5-5.0) NSpRBC administration (units) 8 (5-11) 8 (6-22) NSFFP administration (units) 2 (2-6) 2 (2-12) NSPlatelet administration (bags)† 1 (1) 1 (0-1) NS

pRBC, Packed red blood cells (300 mL); FFP, fresh frozen plasma (300 mL).*Mann-Whitney U test.†One bag of platelet transfusion = 4 pooled units (250 mL).

examine the relationship between ET-1 levels and out-come in 21 patients who required supraceliac and eightwho required infrarenal aortic clamping for repair of nonruptured aortic aneurysm. Unlike the studies byAntonucci et al20 and Fukuda et al,21 plasma ET-1 wasundetectable during the period of aortic clamping and 30minutes after aortic declamping. After 2 hours of reperfu-sion, ET-1 levels were significantly higher in patients inwhom acute kidney failure developed after supraceliacclamping, and at 8 hours, ET-1 levels were significantlyhigher in the supraceliac clamp group compared with theinfrarenal clamp group. In a canine model of infrarenalaortic clamping, Edwards et al23 failed to demonstrate asignificant increase in plasma ET-1 levels during hind limbischemia, but there was a significant increase during reper-fusion and subsequent exsanguination.

In this study, survivors had increased plasma endothe-lin levels during the periods of lower torso ischemia andreperfusion. Furthermore, more than half of the survivorshad increased endothelin levels before operation. Therewas also a significant increase in endothelin levels after 6hours of reperfusion in survivors and nonsurvivors. Theincreased physiological insult of emergency ruptured AAArepair may explain why, unlike previous studies of nonrup-tured AAA repair, elevated endothelin levels were detectedbefore, during, and after operation.

It is interesting to speculate from these data that theET-dependent vasoconstrictor response to hemorrhagicshock, ischemia, and reperfusion has a homeostatic andprotective role in ruptured AAA, in that patients whomanifest a good vasoconstrictor response (which is partlydue to endothelin) have a higher probability of survivalthan those patients whose response is inferior. Thishypothesis would be in keeping with what most vascularsurgeons know intuitively: that is, intense vasoconstriction(as well as “controlled” hypotension, aortic tamponade,and the generation of a prothrombotic state) is one of theprincipal mechanisms that allows patients with rupturedAAA to reach the hospital in better clinical condition andthen undergo successful aneurysm repair.

The reasons for low endothelin levels in nonsurvivors,as well as those patients in whom acute kidney failuredeveloped, are not immediately obvious because rupturedAAA repair is associated with many factors known to stim-ulate endothelin synthesis and secretion: intraoperativehemorrhage and hemodilution, hypoxia, and metabolicacidosis, increased sympathetic discharge and cate-cholamine release, increased cytokine and endotoxinrelease, thrombin generation, and impaired renal excre-tion. In this study, there was no apparent difference in theduration of symptoms of rupture, severity of preoperativeshock, duration of lower torso ischemia or perioperativehematocrit between survivors and nonsurvivors. Mostpatients received renal-dose dopamine in the perioperativeperiod, but, unlike nifedipine, this has not been shown toreduce plasma endothelin levels or maintain glomerular fil-tration rate in patients undergoing major aortic surgery.25

Haynes et al26 reported a significant, and similarly

JOURNAL OF VASCULAR SURGERYVolume 33, Number 6 Adam et al 1245

unexpected, association between high plasma ET-1 levelsand survival in patients with cardiac arrest. They proposedseveral explanations for their findings: poor peripheralblood flow and local tissue acidosis may adversely affectproduction and activity of ET-1, or lead to a local increasein ET-1 that does not enter the circulation; increasednitric oxide production may inhibit endothelin produc-tion; and reduced pulsatile shear stress may lead to selec-tive endothelial cell dysfunction. The low endothelin levelsdemonstrated in nonsurvivors of ruptured AAA and car-diac arrest26 may therefore be an early manifestation ofirreversible whole body hypoperfusion. Studies from thisdepartment have shown that survivors and nonsurvivors ofruptured AAA repair have elevated plasma levels of theendothelial products, tissue plasminogen activator andplasminogen activator inhibitor.27 These and present datalend support to the hypothesis that selective endothelialcell dysfunction may lead to downregulation of theendothelin response in some patients with ruptured AAA,and this may predispose to the development of fatal organdysfunction.

Although most of the current literature concludesthat endothelin release has a pathologic role in critical illness,13-18 the findings of this study do not support thehypothesis that an increased endothelin response predis-poses to poor outcome in patients undergoing rupturedAAA repair. By contrast, elevated perioperative endothelinlevels were associated with survival. Increased circulatingendothelin levels may occur as part of a homeostatic andprotective response to hemorrhage, ischemia, and reperfu-sion in patients with ruptured AAA, or alternatively, lowendothelin levels may be an early marker of severe and irre-versible whole body hypoperfusion in this group of patients.

The authors thank Neil Johnston, Clinical Pharma-cology Unit, University Department of Medical Sciences,

Table II. Postoperative complications

Survivors Nonsurvivors(n = 9) (n = 5)

CardiovascularCardiac failure 4 3Myocardial infarction 1 —Stroke 2 1Critical lower limb ischemia 1 2

RespiratoryRespiratory failure 3 4Acute respiratory distress — 1

syndromePneumonia 6 5

Acute kidney failure 3 5Disseminated intravascular 1 2

coagulationSepsis syndrome 1 2Colon ischemia 1 —Inotropic support

Adrenaline 2 3Dopamine 7 4

JOURNAL OF VASCULAR SURGERY1246 Adam et al June 2001

Western General Hospital, Edinburgh, for technical assis-tance and Amanda Lee, Department of Public HealthSciences, University of Edinburgh, for statistical assistance.

REFERENCES1. Anggard E, Botting R, Vane J. Endothelins. Blood Vessels 1990;27:

269-81.2. Watanabe Y, Naruse M, Monzen C, Naruse K, Ohsumi K, Horiuchi

J, et al. Is big endothelin converted to endothelin-1 in circulatingblood ? J Cardiovasc Pharmacol 1991;17:S503-5.

3. Haynes WG, Webb DJ. Contribution of endogenous generation ofendothelin-1 to basal vascular tone. Lancet 1994;344:852-4.

4. de Nucci G, Thomas R, D’Orleans-Juste P, Antunes E, Walder C,Warner TD, et al. Pressor effects of circulating endothelin are limitedby its removal in the pulmonary circulation and by the release ofprostacyclin and endothelium-derived relaxing factor. Proc Natl AcadSci U S A 1988;85:9797-800.

5. Haynes WG, Webb DJ. The endothelin family of peptides: local hor-mones with diverse roles in health and disease? Clin Sci 1993;84:485-500.

6. Lechleitner P, Genser N, Mair J, Maier J, Artner-Dworzak E, DienstlF, et al. Endothelin-1 in patients with complicated and uncomplicatedmyocardial infarction. Clin Investig 1992;70:1070-2.

7. Cernacek P, Stewart DJ. Immunoreactive endothelin in humanplasma-marked elevations in patients in cardiogenic shock. BiochemBiophys Res Commun 1989;161:562-7.

8. Weitzberg E, Lundberg JM, Rudehill A. Elevated plasma levels ofendothelin in patients with sepsis syndrome. Circ Shock 1991;33:222-7.

9. Sanai L, Haynes WG, MacKenzie A, Grant IS, Webb DJ. Endothelinproduction in sepsis and the adult respiratory distress syndrome.Intensive Care Med 1996;22:52-6.

10. Galatius-Jensen S, Wroblewski H, Emmeluth C, Bie P, Haunso S,Kastrup J. Plasma endothelin in congestive heart failure: a predictor ofcardiac death? J Card Fail 1996;2:71-6.

11. Pacher R, Bergler-Klein J, Globits S, Teufelsbauer H, Schuller M,Krauter A, et al. Plasma big endothelin-1 concentrations in congestiveheart failure with or without systemic hypertension. Am J Cardiol1993;71:1293-9.

12. Tomita K, Ujhe K, Nakanishi T, Tomura S, Matsuda O, Ando K, etal. Plasma endothelin levels in patients with acute renal failure. N EnglJ Med 1989;321:1127.

13. Morise Z, Ueda M, Aiura K, Endo M, Kitajima M. Pathophysiologicrole of endothelin-1 on renal function in rats with endotoxin shock.Surgery 1994;115:199-204.

14. Hemsen A, Modin A, Weitzberg E. Increased concentrations ofendothelin-1 messenger RNA in tissues and endothelin-1 peptide inplasma in septic pigs: modulation by betamethasone. Crit Care Med1996;24:1530-6.

15. Groeneveld AB, Hartemink KJ, de Groot MC, Visser J, Thijs LG.Circulating endothelin and nitrate-nitrite relate to haemodynamic andmetabolic variables in human septic shock. Shock 1999;11:160-6.

16. Wanecek M, Oldner A, Sundin P, Alving K, Weitzberg E, Rudehill A.Effects on haemodynamics by selective endothelin ET (B) receptorand combined endothelin ET (A)/ET (B) receptor antagonism dur-ing endotoxin shock. Eur J Pharmacol 1999;386:235-45.

17. Filep JG. Role for endogenous endothelin in the regulation of plasmavolume and albumin escape during endotoxin shock in conscious rats.Br J Pharmacol 2000;129:975-83.

18. Mitaka C, Hirata Y, Yokoyama K, Nagura T, Tsunoda Y, Amaha K.Pathologic role of endothelin-1 in septic shock. J CardiovascPharmacol 1998;31 Suppl 1:S233-5.

19. Ruetten H, Thiemermann C, Vane JR. Effects of endothelin receptorantagonist, SB 209670, on circulatory failure and organ injury inendotoxic shock in the anaesthetized rat. Br J Pharmacol 1996;118:198-204.

20. Antonucci F, Bertolissi M, Calo I. Plasma endothelin and renal func-tion during infrarenal aortic crossclamping and nifedipine infusion.Lancet 1990;336:1449.

21. Fukuda S, Taga K, Tanaka T, Sakuma K, Fujiwara N, Shimoji K, et al.Relationship between tissue ischemia and venous endothelin-1 duringabdominal aneurysm surgery. J Cardiothorac Vasc Anesth 1995;9:510-4.

22. Lintott P, Berwanger CS, Hafez H, Cheshire N, Manfield AO, WolfeJ, et al. Early endothelin-1 rise predicts renal failure following supra-coeliac clamping. Br J Surg 1998;85:1590-1.

23. Edwards JD, Dovgan PS, Rowley JM, Agrawal DK, Thorpe PE,Adrian TE. Endothelin-1 levels in ischaemia, reperfusion, and haem-orrhagic shock in the canine infrarenal aortic revascularisation model.Eur J Vasc Surg 1994;8:729-34.

24. Rolinski B, Bogner SJ, Goebel FD. Determination of endothelin-1immunoreactivity in plasma, cerebrospinal fluid and urine. Res ExpMed 1994;194:9-24.

25. Antonucci F, Calo L, Rizzolo M, Cantaro S, Bertolissi M, TravagliniM, et al. Nifedipine can preserve renal function in patients undergo-ing aortic surgery with infrarenal crossclamping. Nephron 1996;74:668-73.

26. Haynes WG, Hamer DW, Robertson CE, Webb DJ. Plasma endothe-lin following cardiac arrest: differences between survivors and non-survivors. Resuscitation 1994;27:117-22.

27. Adam DJ, Ludlam CA, Ruckley CV, Bradbury AW. Coagulation andfibrinolysis in patients undergoing operation for ruptured and non-ruptured infrarenal abdominal aortic aneurysm. J Vasc Surg 1999;30:641-50.

Submitted Jun 6, 2000; accepted Nov 3, 2000.