Comparison of Outcomes of CoronaryStenting Versus Conventional Coronary

Angioplasty in the Department ofVeterans Affairs Medical Centers

Charles Maynard, PhD, Steven M. Wright, PhD, Nathan R. Every, MD, andJames L. Ritchie, MD

Although the short-term benefits of stent deploymenthave been established, less is known about long-termoutcomes. This study compares short- and long-termoutcomes in veterans undergoing stenting and conven-tional coronary angioplasty. We used Department ofVeterans Affairs databases to identify 27,224 veteranswho had undergone percutaneous coronary interven-tion (PCI) in Veterans Affairs medical centers betweenOctober 1994 and September 1999. Patients were clas-sified according to whether they had acute myocardialinfarction (AMI) as the principal diagnosis. Baselinecharacteristics were similar in the stent and conventionalgroups. In AMI, hospital mortality was 2.9% for thosewith stents and 4.8% for those who underwent conven-tional coronary angioplasty (p <0.0001), whereas forpaitents without AMI, hospital mortality was similar(1.2% vs 1.4%, p 5 0.12). For AMI, same-admissionbypass surgery rates were lower in the stent group

(0.7% vs 3.2%, p <0.0001) and in the group withoutAMI (1.2% vs 3.3%, p <0.0001). Two-year survival wasbetter for stenting in veterans with (90% vs 88%, p 50.006) and without (92% vs 91%, p 5 0.008) AMI. ForAMI, 2-year rehospitalization rates for PCI (10% vs 13%, p<0.0001), coronary artery bypass surgery (4% vs 6%, p<0.0001), and unstable angina (17% vs 23%) were lowerfor those who had stenting. In the no-AMI group, 2-yearrehospitalization rates for PCI (14% vs 17%, p <0.0001),coronary artery bypass surgery (5% vs 8%, p <0.0001),and unstable angina (22% vs 29%, p <0.0001) werelower in the stent group. Veterans who underwent stentinghad lower hospital mortality, reduced rates of same-ad-mission bypass surgery, marginally better survival, andlower rates of rehospitalization than their counterpartswho had conventional coronary angioplasty. Q2001 byExcerpta Medica, Inc.

(Am J Cardiol 2001;87:1240–1245)

Intracoronary stent implantation has dramatically al-tered the practice of percutaneous coronary inter-

vention (PCI) throughout the world. Clinical trialsfrom the mid-1990s demonstrated that coronary arterystenting resulted in better patient outcomes than didconventional percutaneous transluminal coronary an-gioplasty.1–3 Analysis of much larger and less selec-tive patient populations indicated that stent use wasassociated with significant reductions in the need forsame-admission coronary artery bypass surgery.4–6

The benefits of stenting in the setting of acute myo-cardial infarction (AMI) as well as routine versusprovisional use have also been established.7,8 Recent

results from New York’s Coronary Angioplasty Reg-istry indicated that stent placement was associatedwith lower risk-adjusted long-term mortality, coro-nary artery bypass surgery, and subsequent PCI rates.9

In contrast to this report, most long-term follow-upstudies have included patient populations of hundredsrather than thousands.10–12 In the Department of Vet-erans Affairs (VA) medical centers throughout theUnited States,.5,000 PCIs are performed annually.13

The major purpose of this study was to compare short-and long-term outcomes in veterans undergoing stent-ing with those in veterans receiving conventional bal-loon angioplasty. Given that many veterans are hos-pitalized for acute cardiac conditions in non-VA hos-pitals, data from Medicare will be used to supplementlong-term follow-up information from the VA hospi-tals.14

METHODSStudy population: We used the national VA patient

treatment file to identify patients who received PCIand were discharged from VA medical centers be-tween October 1, 1994, and September 30, 1999. Inall, 33,511 PCIs with International Classification ofDiseases 9th Revision Clinical Modification proce-dure codes 36.01, 36.02, 36.03, 36.05, or 36.06 werechosen. Procedure code 36.06, instituted in October1995, indicates that$1 coronary artery stent was

From the Department of Medicine and Health Services Research andDevelopment, Department of Veterans Affairs, Seattle; Department ofMedicine, Department of Health Services, University of Washington,Seattle, Washington; and Massachusetts Veterans Epidemiology andResearch Information Center, Department of Veterans Affairs BostonHealth Care System, West Roxbury, Massachusetts. This study wassupported by a grant from the VA Quality Enhancement ResearchInitiative Ischemic Heart Disease Center, Seattle, Washington. Manu-script received December 7, 2000; revised manuscript received andaccepted December 27, 2000.

Address for reprints: Charles Maynard, PhD, Department of Vet-erans Affairs, Health Services Research and Development (152),1660 S. Columbian Way, Seattle, Washington 98108. E-mail:cmaynard@u.washington.edu.

The views expressed in this article are those of the authors and do notnecessarily represent the views of the Department of Veterans Affairs.

1240 ©2001 by Excerpta Medica, Inc. All rights reserved. 0002-9149/01/$–see front matterThe American Journal of Cardiology Vol. 87 June 1, 2001 PII S0002-9149(01)01512-0

used. After selecting only the first or index procedure,there were 27,355 veterans available for analysis. Weexcluded veterans who did not have an associatedcardiac diagnosis (diagnosis codes 390 to 429) or whohad procedures in centers performing#5 cases/year,because both of these situations most likely repre-sented coding errors. In the end, 27,224 veteransformed the study population.

Independent variables: The key independent vari-able was whether stenting was performed. Baselinecharacteristics were obtained from the VA patienttreatment file and included age, gender, marital status,race, and year of procedure. Because outcomes inpatients with recent AMI are different from those inpatients with angina pectoris, patients were subse-quently subdivided into 2 distinct biologic groups.The first group included those with a primary (thediagnosis responsible for most of the length of stay) orprincipal (the diagnosis responsible for hospital ad-mission) diagnosis of AMI, and the second containedthose without such a diagnosis.

Comorbidities including hypertension, diabetes,and chronic pulmonary disease were defined fromdiagnosis codes. Also, a comorbidity score as pro-posed by Deyo et al15 was calculated. The Deyo scoreconsiders conditions for the index PCI hospitalization,as well as those#2 years before it. For certain con-ditions (e.g., congestive heart failure, cerebrovascular

disease, AMI as a secondary diagno-sis), the Deyo score is able to distin-guish comorbid conditions from pro-cedural complications.

Dependent variables: In this study,short- and long-term outcomes wereof interest. Short-term or hospitaloutcomes included hospital mortalityand the use of coronary artery bypasssurgery during the same admissionfor PCI. Long-term outcomes, orevents that occurred after hospitaldischarge for the index PCI, werealso considered. In the VA patienttreatment file, vital status after hos-pital discharge was obtained fromthe Beneficiary Identification andRecord Locator System, which is up-dated quarterly and has good reliabil-ity, particularly for veterans who areeligible for cash benefits.16 Vital sta-tus was assessed through March2000. Information about rehospital-ization was obtained from the patienttreatment file for the period October1, 1994, through September 30,1999. Reasons for hospitalizationwere classified using diagnosis andprocedure codes and included (1)PCI (procedure codes 36.01, 36.02,36.05, or 36.06), (2) coronary arterybypass surgery (procedure codes36.11 to 36.19), (3) AMI (diagnosiscode 410.xx), (4) unstable angina

(diagnosis code 411.xx, (5) congestive heart failure(diagnosis code 428.x), and (6) any rehospitalization.

Given that many veterans are hospitalized for acutecardiac conditions in non-VA facilities, we supple-mented follow-up information from the VA with datafrom Medicare provider analysis and review and de-nominator files for calendar years 1996 and 1997. Theprovider analysis and review files contain hospitaldischarge abstract data and the denominator files in-clude eligibility information as well as vital status.Most veterans aged$65 years are eligible for Medi-care on the basis of age, whereas a much smallerproportion of those aged#65 are eligible on the basisof disability. Files from the VA and Medicare werematched by Social Security number, gender, and yearof birth.

Statistical methods: Patient characteristics and out-comes were compared in the stent and no-stent orconventional angioplasty group; the chi-square statis-tic was used for categorical variables and the 2-samplet test for continuous variables.

To adjust for patient differences in the stent andno-stent groups, stepwise logistic regression was usedto compare hospital mortality, same-admission coro-nary artery bypass surgery, and rehospitalization. Todetermine the association between stenting and thevarious measures of outcome, logistic regression wasused in the following way. After all statistically sig-

TABLE 1 Baseline Characteristics of Patients With AMI

VariableStent

(n 5 3,163)No Stent

(n 5 2,909) p Value

Age (yrs) 61 6 11 61 6 11 0.66Year of procedure ,0.0001

1995 0% 36%1996 13% 24%1997 21% 18%1998 28% 13%1999 37% 9%

Men 98% 98% 0.26Marital status 0.04

Divorced 26% 23%Married 56% 58%Never married 7% 6%Separated 5% 5%Unknown 1% 1%Widowed 5% 6%

Race ,0.0001Hispanic white 4% 3%Hispanic black ,1% ,1%American Indian ,1% ,1%Black 10% 12%Asian ,1% ,1%White 79% 80%Unknown 7% 4%

Systemic hypertension 51% 49% 0.09Pulmonary disease 15% 15% 0.96Diabetes mellitus 24% 24% 0.83Deyo score 0.18

0 41% 40%1 34% 34%2 15% 14%3 5% 7%$4 4% 4%

Multivessel angioplasty 11% 6% ,0.0001

CORONARY ARTERY DISEASE/LONG–TERM PCI OUTCOMES IN VETERANS 1241

nificant predictors (p,0.05) listed in Tables 1 and 2were allowed to enter in a stepwise fashion, the stentvariable was forced in to determine its associationwith the specified outcome. The log-rank statistic wasused to compare survival and rehospitalization rates inthe stent and no-stent groups. Stepwise proportional-hazards regression was used to assess the associationbetween stent use and survival, as well as that betweenstent use and rehospitalization.

RESULTSTrends in procedure volumes:

From fiscal year 1995 through fiscalyear 1999, annual angioplasty vol-umes increased 13%, from 6,180 to6,988. The proportion of interven-tions involving stents increased from40% in 1996, to 54% in 1997, to68% in 1998, and to 79% in 1999(Figure 1). The overall number ofstent cases increased 111% from2,610 in 1996, the first year the stentcode was used, to 5,501 in 1999. InVA medical centers, angioplasty vol-umes increased slightly,3%/yearduring the 5-year period, whereas thelarge increase in stent volume aver-aged .27%/year during the 4-yearperiod.

Baseline characteristics: Therewere 27,224 veterans who underwentPCIs during the 5-year period, and22% had a principal or primary di-agnosis of AMI. Overall, stents weredeployed in 49% of veterans, withstent use slightly higher in the groupwith AMI (52% vs 48%). In veteranswith AMI, baseline characteristicsincluding age, gender, marital status,race, hypertension, diabetes, chronicobstructive pulmonary disease, andDeyo score were similar in the stentand conventional angioplasty groups(Table 1). The major difference wasthat more multivessel angioplastywas performed in the stent group.Similar findings were true for thegroup without AMI (Table 2). Al-though many of the comparisonsreached statistical significance due tothe large sample size, the magnitudeof differences was small. As in AMI,more multivessel procedures wereperformed in the stent group.

Hospital outcomes: In the AMIgroup, both unadjusted hospital mor-tality and same-admission bypasssurgery rates were lower in veteransreceiving stents (Table 3). After ad-justing for age, Deyo score, hyper-tension, and year of procedure, stent-ing was still associated with lowerhospital mortality. The strong asso-

ciation between stenting and lower use of same-ad-mission bypass surgery also persisted after adjustmentfor age, year of procedure, and multivessel angio-plasty. In both cases, the unadjusted and adjusted oddsratios were similar.

In the no-AMI group, hospital mortality was sim-ilar in the stent and no-stent groups, and this result didnot change after adjustment for age, Deyo score, andhypertension (Table 3). As in the case with AMI, the

FIGURE 1. Annual stent and conventional coronary angioplasty volumes in Depart-ment of Veterans Affairs medical centers: fiscal years 1995 through 1999.

TABLE 2 Baseline Characteristics in Patients Without AMI

VariableStent

(n 5 10,200)No Stent

(n 5 10,952) p Value

Age (yrs) 62 6 10 62 6 10 ,0.0001Year of procedure ,0.0001

1995 0% 40%1996 17% 24%1997 23% 18%1998 29% 12%1999 32% 8%

Men 99% 99% 0.44Marital status ,0.0001

Divorced 23% 21%Married 60% 62%Never married 6% 5%Separated 4% 5%Unknown 1% 1%Widowed 6% 6%

Race ,0.0001Hispanic white 4% 4%Hispanic black ,1% ,1%American Indian ,1% ,1%Black 8% 10%Asian ,1% ,1%White 82% 81%Unknown 6% 4% 0.001

Systemic hypertension 58% 56% 0.09Pulmonary disease 14% 14% 0.36Diabetes mellitus 28% 28% 0.68Deyo score ,0.0001

0 37% 34%1 35% 34%2 16% 18%3 7% 9%$4 5% 6%

Multivessel angioplasty 12% 8% ,0.0001

1242 THE AMERICAN JOURNAL OF CARDIOLOGYT VOL. 87 JUNE 1, 2001

strong relation between stenting and reduced rates ofbypass surgery was still apparent after adjusting foryear of procedure, chronic pulmonary disease, andmultivessel angioplasty. Again, the unadjusted andadjusted odds ratios were similar.

Long-term survival: The mean and median survivaltimes were both 2.7 years. There were 3,338 totaldeaths during the 5 years; 3,140 were obtained fromVA sources and the remaining 198 (6%) came fromMedicare. In the AMI group, cumulative 2-year sur-vival was 90% for stenting and 88% for conventionalangioplasty (Figure 2), and in the no-AMI group, itwas 92% and 91% for stenting and conventional an-gioplasty, respectively (Figure 3). Differences in sur-vival were statistically significant for both the AMIand no AMI groups (Figures 2 and 3).

Multivariate Cox regression was used to see ifthese differences persisted after adjusting for covari-ates. In the AMI group, the association between stent

use and survival was still evi-dent after adjusting for age,Deyo score, and hypertension(hazard ratio 0.81, 95% confi-dence interval 0.70 to 0.94). Forthe no-AMI group, the associa-tion was not as strong but wasnevertheless statistically signif-icant after adjusting for age,Deyo score, multivessel angio-plasty, hypertension, diabetes,and marital status (hazard ratio0.91, 95% confidence interval0.84 to 0.99).

We also examined with Coxregression the association be-tween stent use and survival forveterans who were dischargedfrom the hospital alive. In theAMI group, improved survivalin the stent group was no longerapparent after adjusting for ageand Deyo score (hazard ratio0.91, 95% confidence interval0.76 to 1.08). In the no-AMIgroup, the same finding wastrue after adjusting for age,Deyo score, marital status, mul-

tivessel angioplasty, and diabetes (hazard ratio 0.92,95% confidence interval 0.84 to 1.01).

Rehospitalization: During the first year after initialhospital discharge, 13,318 veterans were rehospital-ized; only 2.2% of veterans who had repeat hospital-ization were hospitalized in non-VA hospitals. In thegroup with AMI, rehospitalization for any reason dur-ing the first 2 years after initial discharge occurred lessoften in the stent group (Table 4). Specifically, signif-icantly lower proportions in the stent group wererehospitalized for PCI, coronary artery bypass sur-gery, unstable angina, and congestive heart failure.These findings did not change when Cox regressionwas used to adjust for predictors of rehospitalization.In fact, adjusted p values were similar to the unad-justed ones.

In the no-AMI group, the overall pattern of rehos-pitalization favored the stent group (Table 5). In thefirst 2 years after discharge, 66% in the no-stent group

FIGURE 2. Cumulative survival in veterans with and without stents, with AMI.

TABLE 3 Hospital Outcomes

Stent No Stent p ValueUnadjusted OddsRatio (95% CI)

Adjusted OddsRatio (95% CI)

AMIDeath 2.9% 4.8% ,0.0001 0.61 (0.46–0.79) 0.70 (0.51–0.96)Bypass surgery 0.7% 3.2% ,0.0001 0.22 (0.14–0.36) 0.22 (0.13–0.36)No. 3,163 2,909

No AMIDeath 1.2% 1.4% 0.12 0.83 (0.65–1.05) 0.83 (0.66–1.06)Bypass surgery 1.2% 3.3% ,0.0001 0.36 (0.29–0.44) 0.33 (0.26–0.42)No. 10,200 10,952

CI 5 confidence interval.

CORONARY ARTERY DISEASE/LONG–TERM PCI OUTCOMES IN VETERANS 1243

were rehospitalized compared with 55% in the stentgroup (p,0.0001). Two-year cumulative rehospital-ization rates for PCI, coronary artery bypass surgery,unstable angina, and congestive heart failure werelower in the no-stent group. Again, the unadjusteddifferences did not change appreciably after adjust-ment for predictors of rehospitalization.

DISCUSSIONIn this study of veterans undergoing PCI in VA

medical centers from 1995 through 1999, the overallrate of same-admission bypass surgery was 2.2%,markedly less than the 10% figure reported for con-ventional angioplasty in the late 1980s and early1990s.17,18 In addition to the reduced need for coro-nary artery bypass surgery, stenting also was associ-ated with lower hospital mortality in the 6,072 veter-ans with a principal or primary diagnosis of AMI.

Presumably, these short-term benefits are achievedlargely by the ability of the stent to reduce acuteclosures. The results of the present study indicatethat these short-term benefits were sustained in theyears after hospital discharge. The major long-term

outcomes considered in thisstudy were survival and theneed for rehospitalization. Al-though overall 4-year survivalwas about 85% for all veterans,it was statistically better in thestent group. Furthermore,2-year rehospitalization rateswere lower in the stent group.Rehospitalization rates for PCIand bypass surgery were lowerin the stent group, as were re-hospitalization rates for unsta-ble angina and congestive heartfailure. The results of this studyare similar to those reported re-cently by Hannan et al,9 whoreported lower 2-year mortality,PCI, and coronary artery bypasssurgery rates in patients under-going stenting. New York is aspecial case in that only 33 hos-pitals perform PCIs; thereforemost or all of the hospitals arehigh-volume centers with lowermortality and same-admissionbypass surgery rates. Previousstudies have compared 6-month

and 1-year outcomes for stenting andconventional angioplasty, but fewhave the large numbers of patientsand length of follow-up of thepresent study or the Hannanstudy.10–12

Another notable result was thelarge increase in the use of stentingfrom 1995 through 1999. In 1997,54% of procedures in the VA in-cluded coronary artery stents. By

comparison, in a 20% random sample of United Stateshospitals, 59% of procedures performed in 1997 in-volved stenting (unpublished data from 1997 Health-care Cost and Utilization Project). In 1999, almost80% of procedures performed in VA medical centersinvolved stent placement.

One might ask why in 1999 stents were not used in20% of veterans. These veterans may have had diffusedisease or small vessels that precluded the use ofstents. The databases used in this study did not containthe clinical and angiographic data needed to determinewhy these veterans did not undergo stenting. More-over, the databases did not assess restenosis and, inparticular, in-stent restenosis, a most serious compli-cation of stenting.

Another limitation of this study is that follow-upmay not have been complete for all veterans. A rem-edy to this situation was the availability of Medicaredata for 1996 and 1997. Previous studies have shownthat elderly veterans with acute cardiac conditionsoften receive hospital care outside the VA.14 In thepresent study, this was not the case, as only a smallproportion of veterans had Medicare utilization for

FIGURE 3. Cumulative survival in veterans with and without stents, without AMI.

TABLE 4 Cumulative Two-Year Rehospitalization Rates in Patients With AMI

Outcome Stent No StentUnadjusted

p ValueAdjustedp Value

PCI 9.7% 13.3% ,0.0001 ,0.0001Coronary artery bypass surgery 4.3% 6.4% ,0.0001 ,0.0001AMI 14.6% 15.3% 0.52 0.67Unstable angina 17.4% 22.9% ,0.0001 ,0.0001Congestive heart failure 11.1% 13.6% 0.002 0.006Any rehospitalization 49.5% 58.2% ,0.0001 ,0.0001

1244 THE AMERICAN JOURNAL OF CARDIOLOGYT VOL. 87 JUNE 1, 2001

1996 and 1997. There is no comparable database forpersons aged,65 years, so it was not possible toassess non-VA rehospitalization for younger veterans.It does appear, however, that veterans who undergoPCI in VA facilities and require rehospitalization arelikely to be readmitted to VA hospitals. Also, theresults of this study apply to men, because,2% of thestudy population were women. Whether these long-term benefits apply to women is a question worthy offurther investigation.

1. Fischman DL, Leon MB, Baim DS, Schatz RA, Savage MP, Penn I, Detre K,Veltri L, Ricci D, Nobuyoshi M, Cleman M, Heuser R, Almond D, Teirstein PS,Fish RD, Colombo A, Brinker J, Moses J, Shaknovich A, Hirshfield J, Bailey S,Ellis S, Rake R, Goldberg S. A randomized comparison of coronary-stent place-ment and balloon angioplasty in the treatment of coronary artery disease.N EnglJ Med1994;331:496–501.2. Cohen DJ, Krumholz HM, Sukin CA, Ho KKL, Siegrist RB, Cleman M,Heuser RR, Brinker JA, Moses JW, Savage MP, Detre K, Leon MB, Baim DS.In-hospital and one-year economic outcomes after coronary stenting or balloonangioplasty: results from a randomized clinical trial.Circulation 1995;92:2480–2487.3. Versaci F, Gaspardone A, Tomai F, Crea F, Chiariello L Gioffre PA. Acomparison of coronary-artery stenting with angioplasty for isolated stenosis ofthe proximal left anterior descending coronary artery.N Engl J Med1996;334:561–566.4. Ritchie JL, Maynard C, Every NR, Chapko MK. Coronary artery stentoutcomes in a Medicare population: less emergency bypass surgery and lowermortality in stent patients.Am Heart J1999;138:437–440.5. Maynard C, Every NR, Chapko MK, Ritchie JL. Improved outcomes associ-

ated with stenting in the Healthcare Cost and UtilizationProject.J Interven Cardiol; in press.6. Rankin JM, Spinelli JJ, Carere RG, Ricci DR, PennIM, Hilton JD, Henderson MA, Hayden RI, Buller CE.Improved clinical outcome after widespread use of cor-onary artery stenting in Canada.N Engl J Med1999;341:1857–1965.7. Grines CL, Cox DA, Stone GW, Garcia E, MattosLA, Giambartolomei A, Brodie BR, Madonna O, Eij-gelshoven M, Lansky AJ, O’Neill WW, Morice MC.Coronary angioplasty with or without stent implanta-tion for acute myocardial infarction.N Engl J Med1999;341:1949–1956.8. Weaver WD, Reisman MA, Griffin JJ, Buller CE,Leimgruber PP, Henry T, D’Haem C, Clark VL, MartinJS, Cohen DJ, Neil N, Every NR. Optimum percutane-

ous transluminal coronary angioplasty compared with routine stent strategy trial(OPUS-1): a randomised trial.Lancet2000;355:2199–2203.9. Hannan EL, Racz MJ, Arani DT, McCallister BD, Walford G, Ryan TJ. Acomparison of short- and long-term outcomes for balloon angioplasty and coro-nary stent placement.J Am Coll Cardiol2000;36:395–403.10. Matthew V, Grill DE, Scott CG, Garratt KN, Holmes DR. Baseline clinicaland angiographic variables associated with long-term outcome after successfulintracoronary stent implantation.Am J Cardiol1999;84:789–794.11. Betriu A, Masotti M, Serra A, Alonso J, Fernandez-Aviles F, Gimeno F,Colman T, Zueco J, Delcan JL, Garcia E, Calabuig J. Randomized comparison ofcoronary stent implantation and balloon angioplasty in the treatment of de novocoronary artery lesions (START): a four-year follow-up.J Am Coll Cardiol1999;34:1498–1506.12. Kornowski R, Mehran R, Satler LF, Pichard AD, Kent KM, Greenberg A,Mintz GS, Hong MK, Leon MB. Procedural results and late clinical outcomesfollowing multivessel stenting.J Am Coll Cardiol1999;33:420–426.13. Ritchie JL, Maynard C, Chapko MK, Every NR, Martin DC. A comparisonof percutaneous transluminal coronary angioplasty in the Department of VeteransAffairs and in the private sector in the state of Washington.Am J Cardiol1998;81:1094–1099.14. Wright SM, Petersen LA, Lamkin R, Daley J. Increasing use of Medicareservices by veterans with acute myocardial infarction.Med Care1999;37:529–537.15. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index foruse with ICD-9-CM administrative databases.J Clin Epidemiol1992;45:613–619.16. Dominitz JA, Maynard C, Boyko EJ. Assessment of vital status in Depart-ment of Veterans Affairs National Databases: comparison with state death cer-tificates.Ann Epidemiol; in press.17. King SB, Lembo NJ, Weintraub WS, Kosinski AS, Barnhart HX, Kutner MH,Alazraki NP, Guyton RA, Zhao XQ. A randomized trial comparing coronaryangioplasty with coronary bypass surgery.N Engl J Med1994;331:1044–1050.18. The Bypass Angioplasty Revascularization (BARI) Investigators. Compari-son of coronary bypass surgery with angioplasty in patients with multivesseldisease.N Engl J Med1996;335:217–225.

TABLE 5 Cumulative Two-Year Rehospitalization Rates in Patients Without AMI

Outcome Stent No StentUnadjusted

p ValueAdjustedp Value

PCI 14.1% 16.7% ,0.0001 ,0.0001Coronary artery bypass surgery 5.0% 7.7% ,0.0001 ,0.0001AMI 7.0% 7.3% 0.24 0.21Unstable angina 22.1% 28.7% ,0.0001 ,0.0001Congestive heart failure 11.8% 12.7% 0.002 0.053Any rehospitalization 54.6% 65.9% ,0.0001 ,0.0001

CORONARY ARTERY DISEASE/LONG–TERM PCI OUTCOMES IN VETERANS 1245